[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M6WWM5J4TXYYK1DBZ9WRNWJG":3,"skill-same:01M6WWM5J4TXYYK1DBZ9WRNWJG":252},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":16,"categories":17,"keywords":28,"publishedAt":49,"linkOnly":50,"markdownContent":51,"markdownBody":52,"frontMatter":53,"refName":80,"commitSha":81,"skillDirectory":82,"forksCount":14,"repositoryHtmlUrl":83,"repositoryHomepageUrl":84,"ownerHtmlUrl":85,"ownerBlogUrl":6,"ownerDescription":6,"updatedAt":86,"install":87,"resources":100},"01M6WWM5J4TXYYK1DBZ9WRNWJG","pk-doctor",null,"Aggregator health-check for a processkit-managed repository. Runs a fixed suite of checks over the live context\u002F tree plus src\u002Fcontext\u002F drift, then resolves ERROR \u002F WARN \u002F actionable INFO findings by default. Use when the user invokes `\u002Fpk-doctor`, asks \"is this repo healthy\", or has just finished an upgrade and wants a sanity pass. Info-only mode is opt-in: stop after the report only when the user explicitly says info-only, dry-run, report-only, check-only, or no fixes.","projectious-work\u002Fprocesskit","projectious-work","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F263889726?v=4","01M6N97TJJ118WQ35CHRV8RD0E","projectious.work","01M6N97TPWJPF0BMKYQ4PBY5W8",0,"MIT","https:\u002F\u002Fgithub.com\u002Fprojectious-work\u002Fprocesskit\u002Fblob\u002Fmain\u002Farchive\u002Fprocesskit-install-20261101\u002Fcontext\u002Fskills\u002Fprocesskit\u002Fpk-doctor\u002FSKILL.md",[18,23],{"categoryNo":19,"categoryName":20,"taxonomyType":21,"taxonomyTypeDesc":22},"01JBSEEDCAT000000000000023","Skill与智能体开发","SCENARIO","使用场景",{"categoryNo":24,"categoryName":25,"taxonomyType":26,"taxonomyTypeDesc":27},"01JBSEEDCAT000000000000034","行政与办公人员","PROFESSION","适用职业",[29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48],"pk doctor","doctor","dry","when","phase","managed","mode","errors","sanity","actionable","error","aggregator","upgrade","info","drift","health-check","default","log","only","fix","2026-11-10 10:12:30",false,"---\nname: pk-doctor\ndescription: |\n  Aggregator health-check for a processkit-managed repository. Runs a fixed suite of checks over the live context\u002F tree plus src\u002Fcontext\u002F drift, then resolves ERROR \u002F WARN \u002F actionable INFO findings by default. Use when the user invokes `\u002Fpk-doctor`, asks \"is this repo healthy\", or has just finished an upgrade and wants a sanity pass. Info-only mode is opt-in: stop after the report only when the user explicitly says info-only, dry-run, report-only, check-only, or no fixes.\nmetadata:\n  processkit:\n    apiVersion: processkit.projectious.work\u002Fv2\n    id: SKILL-pk-doctor\n    version: \"1.0.0\"\n    created: 2026-04-20T00:00:00Z\n    category: processkit\n    layer: 3\n    uses:\n      - skill: event-log\n        purpose: Emit the single `doctor.report` LogEntry per run (the audit trail for every invocation).\n      - skill: migration-management\n        purpose: List pending migrations for the `migrations` check and apply them under `--fix`.\n      - skill: index-management\n        purpose: (future) surface index errors alongside doctor findings in Phase 2.\n    commands:\n      - name: pk-doctor\n        args: \"[info-only | dry-run | --category=... | --fix=... | --fix-all | --since=\u003Cref> | --yes]\"\n        description: \"Run health checks and resolve actionable findings by default.\"\n    provides:\n      mcp_tools:\n        - run_pk_doctor\n---\n\n# pk-doctor\n\n## Intro\n\n`\u002Fpk-doctor` is the single user-visible health-check surface for a\nprocesskit repository. It is the diagnostic and remediation entry point\nyou run after an upgrade, before a release, or any time you need to\nanswer the question \"is this repo healthy?\" in one command. The skill\npattern is deliberately modelled on `npm doctor`, `brew doctor`, and\n`rustup doctor`, with one processkit-specific default: agents resolve\nfindings by default and stop at report-only mode only when the user\nexplicitly asks for info-only\u002Fdry-run output.\n\n### Container\u002Fhost boundary\n\n`\u002Fpk-doctor`, `run_pk_doctor`, and `doctor.py` are the derived-project\nhealth surfaces for the running development container. In aibox-derived\nprojects, agents must use these processkit surfaces from inside the\ncontainer for processkit repository, MCP, migration, context, and\nruntime-health checks.\n\nHost-side installer diagnostics belong to the owner\u002Foperator outside the\ncontainer. pk-doctor skill text, command text, MCP descriptions, and\nfindings must not direct an in-container agent to run host-side\norchestrator health checks. When a finding requires host installer repair\nsuch as sync or rebuild, phrase it as an explicit host action for the\nowner and keep the processkit-side fix generic.\n\nSeverity and actionability are separate contracts. `ERROR`\u002F`WARN`\u002F`INFO`\ndescribe blocking severity; `action_required`, `action_kind`,\n`default_agent_action`, `requires_user_confirmation`, and\n`acceptable_resolution` tell agents what must happen next. A severity\ndowngrade is not a resolution. Clean means no blocking findings and no\nunresolved actionable findings, or a durable disposition for every\nactionable finding.\n\nThis skill is the Phase 1 landing. It started with four checks. Phase 2 and\nPhase 3 add more checks, lift the aggregator into an MCP server, and\nharden the `--fix` paths — each with their own WorkItem.\n\n## Overview\n\n### Invocation\n\n```\n\u002Fpk-doctor                              # check, then resolve actionable findings\n\u002Fpk-doctor info-only                    # report only; do not resolve\n\u002Fpk-doctor dry-run                      # synonym for info-only\n\u002Fpk-doctor --category=schemas,migrations\n\u002Fpk-doctor --fix=migrations             # script-level scoped fixer\n\u002Fpk-doctor --fix-all                    # script-level fixers only; still review output\n\u002Fpk-doctor --since=v0.18.2              # scope file-walk checks only\n\u002Fpk-doctor --yes                        # non-interactive (auto-confirms safe fixes)\n```\n\nThe MCP wrapper accepts the same confirmation explicitly as\n`run_pk_doctor(fix=\"...\", yes=true)`. Calls without `yes=true` remain\ndetect-only when a category requires confirmation.\n\nThe script flags `--fix` and `--fix-all` are mutually exclusive. Exit\ncode is `0` if no ERRORs were produced, `1` otherwise. Every script run\n— regardless of outcome — writes exactly one `doctor.report` LogEntry\nvia the event-log MCP.\n\n### Default remediation workflow\n\nWhen a user invokes `\u002Fpk-doctor` or asks to check repository health, the\nagent must treat remediation as the default:\n\n1. Run `run_pk_doctor()` or the local `doctor.py` report.\n2. Build a queue of every `ERROR`, every `WARN`, and every finding whose\n   `action_required` is true, including actionable INFO findings.\n3. Resolve safe local fixes immediately through the owning script,\n   command, or MCP tool. Use the reported `fix_mcp_tool`,\n   `suggested_fix`, `action_kind`, and `default_agent_action` fields to\n   choose the narrowest correction.\n4. For migration findings, route through migration-management. Apply or\n   continue unambiguous migrations; reject malformed no-op migrations\n   only when the finding itself identifies the defect and a rejection\n   reason is clear. Ask the user only when the migration changes policy,\n   deletes data, has unclear intent, or conflicts with local work.\n5. For archive, policy, external-dependency, or user-confirmation\n   findings, either resolve them if the fix is already known and\n   non-destructive, create\u002Flink tracking when the finding calls for\n   tracking, or ask a concise question when a human decision is required.\n6. Rerun pk-doctor after remediation and report the final ERROR\u002FWARN and\n   action totals.\n\nIf the user explicitly says \"info only\", \"dry run\", \"report only\",\n\"check only\", \"no fixes\", or equivalent, stop after the report and do\nnot mutate anything. In that mode, still highlight the concrete next\nactions so the report is actionable.\n\n### The four Phase 1 checks\n\n1. **`schema_filename`** — walks every `context\u002F\u003Ckind>\u002F**\u002F*.md`. Loads\n   the matching schema from `src\u002Fcontext\u002Fschemas\u002F\u003Ckind>.yaml` (the LIVE\n   source, never the template mirror) and validates the file's frontmatter\n   against it. Also checks that the filename stem equals `metadata.id`\n   and that any date encoded in the filename matches `metadata.created`.\n   Phase 1 is WARN-only for rename suggestions — no auto-fix.\n2. **`sharding`** — logs must live under `context\u002Flogs\u002FYYYY\u002FMM\u002F` subdirs;\n   migrations must live under their `spec.state` subdir (`pending\u002F`,\n   `in-progress\u002F`, `applied\u002F`). WARN on any file in the wrong bucket.\n3. **`migrations`** — calls\n   `mcp__processkit-migration-management__list_migrations(state=\"pending\")`.\n   INFO for the pending count; WARN for any pending entry older than\n   14 days. Under `--fix=migrations` or `--fix-all`, interactively\n   prompts `Apply MIG-xxx? [y\u002FN\u002Fs]` per entry and invokes\n   `apply_migration` on yes.\n4. **`drift`** — subprocess-invokes `scripts\u002Fcheck-src-context-drift.sh`\n   from the repo root. Exit 0 → one INFO \"trees in sync\". Exit 1 →\n   one WARN per offending line reported by the script.\n5. **`commands_consistency`** — walks every\n   `context\u002Fskills\u002F*\u002F*\u002FSKILL.md` and verifies that each entry in\n   `metadata.processkit.commands:` has a matching `commands\u002F\u003Cname>.md`\n   file alongside the SKILL.md, uses the reserved `pk-` prefix, and has\n   matching `argument-hint` frontmatter. When `.claude\u002Fcommands\u002F` or\n   `.agents\u002Fskills\u002F` exists, it also verifies those harness projections\n   exactly match the canonical command set.\n\nAdditional checks added after Phase 1:\n\n- **`team_consistency`** — wraps `team-manager.check_all()`.\n- **`team_member_exports`** — reconciles active TeamMembers under `context\u002Fteam-members\u002F` with Claude sub-agent export files under `.claude\u002Fagents\u002F\u003Cslug>.md` (the path emitted by `team-manager.export_claude_subagent`). Emits WARN per active TeamMember missing an export (harness can't dispatch as `subagent_type`) and WARN per stale export whose slug no longer maps to an active TeamMember; INFO when the roster and the export dir are in sync. Detect-only — fix is to re-run `export_claude_subagent` (or `export_claude_subagents`) or delete the stale file. Lands the harness signal half of the sub-agent-dispatch clause from BACK-20260509_1317-WildPanda.\n- **`release_integrity`** — see DEC-20260422_1348-SnowyWolf-local-only-release-bulletproofing.\n- **`mcp_config_drift`** — reads `context\u002F.processkit-mcp-manifest.json`\n  (produced at release time by `scripts\u002Fgenerate-mcp-manifest.py`) and\n  recomputes the per-skill `mcp-config.json` sha256es. A missing manifest\n  or stale aggregate surfaces as WARN (run the generator). In a\n  derived-project context (`aibox.lock` + `.mcp.json` both present at\n  the repo root), any processkit server missing from `.mcp.json`'s\n  `mcpServers` map surfaces as ERROR with an explicit host-action hint\n  asking the owner to re-run the project installer\u002Fsync tool outside the\n  container. Exists because some host installers gate `.mcp.json`\n  re-merge on processkit version delta only, so per-skill MCP-config\n  edits within a release cycle may not reach derived projects until the\n  next version bump — see DEC-20260423_2049-VastLake.\n- **`server_header_drift`** — walks every `context\u002Fskills\u002Fprocesskit\u002F*\u002Fmcp\u002Fserver.py`,\n  hashes its PEP 723 inline metadata block (`# \u002F\u002F\u002F script` ... `# \u002F\u002F\u002F`),\n  and compares against the manifest's `per_server_header` baseline. WARN\n  on any drift listing the affected skill slugs — signal that someone\n  edited a dep header without regenerating the manifest, and the harness\n  needs a restart to pick up the new deps. Detect-only; the fix is\n  user-driven (regenerate manifest + restart harness). Per\n  DEC-20260424_0127-QuickPine (SharpBrook split, RapidSwan ships the\n  dep-drift half).\n- **`preauth_applied`** — compares\n  `context\u002Fskills\u002Fprocesskit\u002Fskill-gate\u002Fassets\u002Fpreauth.json` against\n  `.claude\u002Fsettings.json` and `.codex\u002Fconfig.toml`. WARNs when aibox has\n  not merged processkit MCP preauthorization into either harness config,\n  which is the common cause of reauthorization prompts after container\n  rebuilds.\n- **`context_consumption`** — emits an INFO-only estimate of processkit\n  context footprint by group (`startup`, command adapters, skill docs,\n  MCP configs) using `ceil(utf8 bytes \u002F 4)` as the provider-neutral\n  token heuristic. It is intentionally not an MCP tool, so measurement\n  does not add another tool schema to the harness context.\n  It also exposes a local checkpoint\u002Freport CLI:\n  ```sh\n  python context\u002Fskills\u002Fprocesskit\u002Fpk-doctor\u002Fscripts\u002Fchecks\u002Fcontext_consumption.py checkpoint \u003Clabel>\n  python context\u002Fskills\u002Fprocesskit\u002Fpk-doctor\u002Fscripts\u002Fchecks\u002Fcontext_consumption.py report \u003Cbefore> \u003Cafter>\n  ```\n  Checkpoints are JSON files under\n  `context\u002F.state\u002Fcontext-consumption\u002Fcheckpoints\u002F`. Reports compare\n  observed processkit payloads and label token counts as local estimates,\n  not provider-billed usage.\n- **`v2_contracts`** — validates v2 API contract invariants across WorkItem, Binding, Artifact, and LogEntry entities. Catches process-instance definitions missing their definition reference, time-window bindings without a recurrence rule, orphaned cost-policy artifacts (not bound to a budget), policy supersession chains with breaks, uncalibrated eval-spec judges, and stale or missing agent-card projections. Emits ERROR for all contract violations; detect-only.\n- **`context_hygiene`** — validates artifact naming policies (model-spec and model-profile use timestamped `ART-YYYYMMDD_HHMM-*` scheme), model binding integrity (role\u002FTeamMember defaults must target provider-neutral model-profile artifacts unless marked direct_model_pin), and cross-reference health. Also detects demoted schema kinds still present in src\u002F, detects archive candidates, warns on mixed binding filename styles, and checks sqlite-vec semantic index health. Emits WARN \u002F ERROR depending on severity; detect-only.\n- **`sensitive_data` allowlist** — intentional public or synthetic email\n  addresses may be listed in `.pk-doctor-allowlist.toml`:\n  ```toml\n  [sensitive_data.email_allowlist]\n  addresses = [\"info@projectious.work\", \"deploy@example.local\"]\n  ```\n  The allowlist applies only to email-address findings; secrets and other PII\n  are never suppressed.\n- **`runtime_health`** — owns in-container runtime probes that host-side\n  installer diagnostics cannot verify directly. Checks lnav availability\n  for the Prefix L structured log viewer, sqlite-vec import\u002Fload health\n  in the pk-doctor\u002FMCP runtime, Codex bubblewrap sandbox smoke status\n  when Codex is enabled, PID 1 sleep-infinity hygiene, cgroup\n  memory\u002Fpid\u002FOOM pressure, processkit Python MCP process counts,\n  PowerKit image\u002Fplugin tree, status plugin scripts, local render\n  helpers, and runtime-home write probes. Emits WARN only for\n  container-local actionable failures; host-only runs emit a skip INFO.\n- **`entity_storage_hygiene`** — validates local `context\u002F`\n  storage policy separately from template freshness: unmanaged host\n  artifacts, demoted legacy roots such as `context\u002Fmodels\u002F`,\n  root-level migration briefings, mixed root\u002Fsharded entity layouts,\n  mixed filename policies, placeholder `0000` timestamps, and\n  TeamMember private\u002Fslug policy explanations. Emits WARN for\n  actionable storage drift; legacy layouts and mixed filename policies\n  are migration work. `storage.root-migration-briefings` can be\n  auto-resolved via `--fix=entity_storage_hygiene` by archiving\n  non-completed CLI migration briefings.\n- **`schema_vocabulary`** — validates closed-vocabulary subtype fields across v2 entities by comparing frontmatter values against Schema-declared known-kinds\u002Fknown-types (Artifact kind, Binding type, WorkItem type, LogEntry event_type, Migration kind). `legacy_known_*` schema fields are not accepted as a terminal state; legacy values require an explicit data-fix Migration or a schema migration. Also validates that Migration entities include required v2 version metadata fields. Emits ERROR for unknown vocabulary values or missing version fields; detect-only.\n- **`migration_integrity`** — flags malformed pending migrations that exhibit the \"empty-baseline\" defect (same-version migrations with content in affected_groups\u002Faffected_files despite no-op intent, or affected_groups populated but affected_files empty). Per BACK-20260425_1711-CleverRiver, both patterns are defect signatures. Emits WARN with suggested fix to reject via migration-management MCP and refile the upstream bug; detect-only.\n- **`mcp_gateway`** — reports processkit-gateway MCP config health and harness registration state. Validates gateway config presence\u002Fstructure at `context\u002Fskills\u002Fprocesskit\u002Fprocesskit-gateway\u002Fmcp\u002Fmcp-config.json`, checks env vars and command launch target, detects mixed gateway+granular server registrations in `.mcp.json`, and warns on insecure daemon bindings. Emits INFO on healthy state, WARN on minor gaps (missing proxy --url, nonlocal daemon bind), ERROR on parsing failure; detect-only.\n- **`skill_dag`** — builds the skill-dependency directed-acyclic-graph (DAG) from `metadata.processkit.uses[*].skill` entries in every `context\u002Fskills\u002F**\u002FSKILL.md` and validates references exist, graph is acyclic, and layer constraints are honored (skill layer N only uses skills with layer ≤ N). Emits ERROR for missing references, cycles (with full path), or layer violations; detect-only.\n- **`v1_entity_drift`** — walks every registered v2 entity directory under `context\u002F` and surfaces files whose frontmatter still declares `apiVersion: …\u002Fv1`. Emits WARN per file with a hardcoded successor table (Actor→TeamMember, Process→Scope+Gate, StateMachine→lifecycle metadata, Model→Artifact(model-spec)); v1 files in append-only buckets (`context\u002Flogs\u002F`, `context\u002Fmigrations\u002Fapplied\u002F`) are downgraded to a single INFO since they are intentionally historical. Under `--fix=v1_entity_drift` (or `--fix-all`) the check interactively records `apply_migration` intents when a pending v1->v2 Migration already covers the target, otherwise records `propose_migration` intents — it never auto-creates migrations or hand-edits files. Per BACK-20260509_1318-KindSpruce.\n- **`agents_md_hygiene`** — validates the derived-project startup\n  policy surface. Checks root `AGENTS.md` for `pk-managed:*` blocks\n  (`pk-compliance-contract-v2`, `pk-commands`), compares managed block\n  content against `src\u002FAGENTS.md` or the latest template when available,\n  verifies processkit references for session start, skill routing,\n  entity IO, migrations, decision capture, sub-agent dispatch,\n  TeamMember\u002Fmodel binding, MCP topology, provider pointer files, and\n  command blocks, and flags stale guidance such as legacy actor roots,\n  `spec.x_aibox`, model-tier role tables, or duplicated provider-specific\n  policy. WARN findings include a `briefing` payload for a project agent\n  to reconcile the file after processkit upgrades. Detect-only; managed\n  block replacement and local policy merges remain user-reviewed.\n- **`sensitive_data`** — scans tracked text\u002Fconfig files for\n  deterministic high-signal secret and personal-data patterns: private\n  key blocks, provider tokens, JWTs, URL credentials, assigned\n  high-entropy secret values, email addresses, phone numbers, SSN-like\n  values, and credit-card-like numbers with a valid Luhn checksum.\n  Deterministic secret findings are ERROR\u002FWARN and require human\n  confirmation because the correct response may be rotation, redaction,\n  migration, or explicit policy acceptance. The check also emits one\n  INFO `sensitive-data.probabilistic-briefing` that tells the\n  derived-project agent what regex cannot prove: real names and aliases,\n  addresses, dates of birth, customer\u002Faccount IDs, medical\u002Ffinancial\u002F\n  employment data, screenshots, archives, database dumps, logs, and\n  non-standard short secrets. The briefing includes non-triggering\n  deterministic and probabilistic example prompts so an agent knows what\n  to search for without adding realistic secrets to the shipped scanner.\n  This is an advisory layer, not a complete DLP engine.\n- **`supply_chain`** — performs offline supply-chain hygiene checks for\n  dependency manifests\u002Flockfiles, license policy, and local scanner\n  outcomes:\n  - emits `ERROR` on denied licenses, missing application lockfiles, and\n    high\u002Fcritical vulnerabilities;\n  - emits `WARN` on unknown\u002Freview\u002Fallowed-policy misses and skipped\n    security scanners;\n  - emits `INFO` for inventory counts and SBOM discoveries.\n  Security\u002Foutdated\u002Fsupplier-quality checks stay advisory and require\n  explicit core-surface opt-in data; pk-doctor performs no outbound\n  network calls during this check.\n\n### What doctor will NEVER do\n\n- Hand-edit any file under `context\u002F`. All writes route through the\n  appropriate MCP tool (e.g. `apply_migration`).\n- Touch `context\u002Ftemplates\u002F` (read-only diff baseline).\n- Touch `.mcp.json` (merged file; use the generated per-skill\n  `mcp\u002Fmcp-config.json` and gateway catalog instead).\n- Write to `context\u002Flogs\u002F` directly — reports go through\n  `mcp__processkit-event-log__log_event`.\n- Treat a clean severity tally as done while actionable INFO findings\n  remain unresolved.\n- Stop after the report unless the user explicitly asked for info-only,\n  dry-run, report-only, check-only, or no-fix behavior.\n\n### Report shape\n\nAfter the stdout human-readable summary, the skill emits a single\n`doctor.report` LogEntry:\n\n```json\n{\n  \"event_type\": \"doctor.report\",\n  \"summary\": \"\u002Fpk-doctor --fix=migrations — 0 ERROR \u002F 2 WARN \u002F 3 INFO\",\n  \"details\": {\n    \"doctor_version\": \"1.0.0\",\n    \"invocation\": \"\u002Fpk-doctor --fix=migrations\",\n    \"categories\": {\n      \"schema_filename\": {\"ERROR\": 0, \"WARN\": 1, \"INFO\": 140},\n      \"sharding\": {\"ERROR\": 0, \"WARN\": 0, \"INFO\": 1},\n      \"migrations\": {\"ERROR\": 0, \"WARN\": 1, \"INFO\": 1},\n      \"drift\": {\"ERROR\": 0, \"WARN\": 0, \"INFO\": 1}\n    },\n    \"action_totals\": {\n      \"actionable\": 2,\n      \"needs_user_confirmation\": 1,\n      \"needs_tracking\": 0,\n      \"safe_fix\": 1,\n      \"migration_needed\": 1,\n      \"archive_needed\": 0,\n      \"policy_decision_needed\": 0,\n      \"external_dependency\": 0\n    },\n    \"top_findings\": [\n      {\n        \"severity\": \"WARN\",\n        \"id\": \"migration.stale-pending\",\n        \"entity_ref\": \"MIG-...\",\n        \"message\": \"...\",\n        \"action_required\": true,\n        \"action_kind\": \"migration_needed\",\n        \"default_agent_action\": \"create_migration\",\n        \"requires_user_confirmation\": true,\n        \"acceptable_resolution\": \"migrated\"\n      }\n    ],\n    \"fixes_applied\": [],\n    \"duration_ms\": 2340\n  }\n}\n```\n\n`top_findings` caps at 20 to keep the log entry compact; the full\nstdout report remains authoritative for the details of a specific run.\nThe structured JSON payload returned by `run_pk_doctor` includes the\nsame `action_totals` and actionability fields on every finding.\n\n### Interactive prompts\n\nInteractive prompts gate on `sys.stdin.isatty()`. If a `--fix` was\nrequested but the process isn't attached to a terminal, the skill emits\na `WARN` (`fix.non-interactive`) and skips the fix — it does not silently\nprompt and hang. `--yes` auto-confirms every prompt except any marked\n`data_loss=True` (none in Phase 1, but the hook is present).\n\n### `--since` scope\n\n`--since=\u003Cgit-ref>` restricts only the file-walk checks\n(`schema_filename`, `sharding`). The `migrations` and `drift` checks\nalways run full-scan: migrations pull from an MCP tool with its own\nstate, and drift compares entire trees.\n\n### Resolved policy defaults\n\nThese were agreed with the owner during shape review:\n\n- **YAML-datetime vs JSON-Schema string coercion** failures surface as\n  `WARN schema.datetime-coercion` (not ERROR). Parser-layer quirk.\n- **Command projections** — canonical command adapters live under\n  `context\u002Fskills\u002F**\u002Fcommands\u002F`; slash-capable harness projections live\n  under `.claude\u002Fcommands\u002F`; natural-language command shims live under\n  `.agents\u002Fskills\u002F`. `src\u002Fcontext\u002F` mirrors only the canonical source.\n- **Filename auto-rename** — WARN only in Phase 1. Phase 2 adds\n  orchestrated rename + reindex.\n\n## Gotchas\n\nAgent-specific failure modes — provider-neutral pause-and-self-check items:\n\n- **Confusing remediation mode with blind `--fix-all`.** The skill is\n  remediation-first by default, but that does not mean `doctor.py\n  --fix-all --yes` is safe. First run the report, read actionability,\n  then apply the narrowest known fix through the owning tool. Ask only\n  when a finding is destructive, policy-sensitive, external, or unclear.\n- **Stopping after a report when the user did not ask for info-only.**\n  `\u002Fpk-doctor` means \"check and clean up what can be cleaned up.\" If the\n  user only wants a briefing, they must say info-only, dry-run,\n  report-only, check-only, or no fixes.\n- **Treating drift warnings as noise.** The drift check exists because\n  v0.15.0–v0.18.0 shipped four releases with silent drift. If the\n  drift check produces a WARN, it has found a real divergence —\n  do not dismiss without investigating.\n- **Hand-editing files to \"clear\" a doctor warning.** Doctor warnings\n  are diagnostic. Fixing the underlying cause through the right MCP\n  write tool (not by editing the offending file directly) preserves\n  the audit trail. A WARN that reappears next run is more useful than\n  a hand-fixed tree with no record of the fix.\n- **Validating against `context\u002Ftemplates\u002F` schemas.** The template\n  mirror is a diff baseline, not the live schema source. Validation\n  must read `src\u002Fcontext\u002Fschemas\u002F\u003Ckind>.yaml` — the authoritative\n  copy. Validating against templates will produce stale false\n  positives after any schema evolution.\n- **Forgetting the `doctor.report` LogEntry on dry-run.** Even a\n  detect-only run must emit the report: it is the audit record of the\n  invocation. A run that prints to stdout but logs nothing strands\n  the next session with no record of what was checked.\n- **Running doctor mid-migration.** Doctor's `migrations` check will\n  list pending migrations and (under `--fix`) offer to apply them. If\n  you are already in the middle of a manual migration review, running\n  `--fix=migrations` can reorder work silently. Finish or pause the\n  manual review first.\n\n## Full reference\n\n### CLI contract\n\n```\ndoctor.py [--category=LIST] [--fix=LIST | --fix-all] [--since=REF] [--yes]\n```\n\n| Flag           | Meaning |\n|----------------|---------|\n| `--category`   | Comma list of categories to run. Default: all four. |\n| `--fix`        | Comma list of categories to enable fixes for. Mutex with `--fix-all`. |\n| `--fix-all`    | Enable fixes for every category that supports them. Mutex with `--fix`. |\n| `--since`      | Git ref; restricts `schema_filename` + `sharding` to files changed since. |\n| `--yes`        | Auto-confirm all non-`data_loss` fix prompts. |\n\n### Adding a check in Phase 2\n\n1. Drop a new module in `scripts\u002Fchecks\u002F\u003Cname>.py` exporting a\n   `run(ctx) -> list[CheckResult]` function (and optional\n   `run_fix(ctx, results) -> list[dict]`).\n2. Register it in `scripts\u002Fchecks\u002F__init__.py` in `REGISTRY`.\n3. Add a gotcha for that check here in SKILL.md.\n4. Mirror both files into `src\u002Fcontext\u002Fskills\u002Fprocesskit\u002Fpk-doctor\u002F`.\n\nThe aggregator in `doctor.py` does not need to change.\n\n### What this skill does NOT do (yet)\n\n- Validate cross-references (Phase 2).\n- Check `PROVENANCE.toml` drift (Phase 2).\n- Check `mcp-config.json` integrity (Phase 2).\n- Smoke-import MCP servers (Phase 2).\n- Detect version skew between `aibox.lock` and installed SKILL.md pins\n  (Phase 2).\n- Expose itself as an MCP server for machine invocation (Phase 3, only\n  if demand shows up).\n","# pk-doctor\n\n## Intro\n\n`\u002Fpk-doctor` is the single user-visible health-check surface for a\nprocesskit repository. It is the diagnostic and remediation entry point\nyou run after an upgrade, before a release, or any time you need to\nanswer the question \"is this repo healthy?\" in one command. The skill\npattern is deliberately modelled on `npm doctor`, `brew doctor`, and\n`rustup doctor`, with one processkit-specific default: agents resolve\nfindings by default and stop at report-only mode only when the user\nexplicitly asks for info-only\u002Fdry-run output.\n\n### Container\u002Fhost boundary\n\n`\u002Fpk-doctor`, `run_pk_doctor`, and `doctor.py` are the derived-project\nhealth surfaces for the running development container. In aibox-derived\nprojects, agents must use these processkit surfaces from inside the\ncontainer for processkit repository, MCP, migration, context, and\nruntime-health checks.\n\nHost-side installer diagnostics belong to the owner\u002Foperator outside the\ncontainer. pk-doctor skill text, command text, MCP descriptions, and\nfindings must not direct an in-container agent to run host-side\norchestrator health checks. When a finding requires host installer repair\nsuch as sync or rebuild, phrase it as an explicit host action for the\nowner and keep the processkit-side fix generic.\n\nSeverity and actionability are separate contracts. `ERROR`\u002F`WARN`\u002F`INFO`\ndescribe blocking severity; `action_required`, `action_kind`,\n`default_agent_action`, `requires_user_confirmation`, and\n`acceptable_resolution` tell agents what must happen next. A severity\ndowngrade is not a resolution. Clean means no blocking findings and no\nunresolved actionable findings, or a durable disposition for every\nactionable finding.\n\nThis skill is the Phase 1 landing. It started with four checks. Phase 2 and\nPhase 3 add more checks, lift the aggregator into an MCP server, and\nharden the `--fix` paths — each with their own WorkItem.\n\n## Overview\n\n### Invocation\n\n```\n\u002Fpk-doctor                              # check, then resolve actionable findings\n\u002Fpk-doctor info-only                    # report only; do not resolve\n\u002Fpk-doctor dry-run                      # synonym for info-only\n\u002Fpk-doctor --category=schemas,migrations\n\u002Fpk-doctor --fix=migrations             # script-level scoped fixer\n\u002Fpk-doctor --fix-all                    # script-level fixers only; still review output\n\u002Fpk-doctor --since=v0.18.2              # scope file-walk checks only\n\u002Fpk-doctor --yes                        # non-interactive (auto-confirms safe fixes)\n```\n\nThe MCP wrapper accepts the same confirmation explicitly as\n`run_pk_doctor(fix=\"...\", yes=true)`. Calls without `yes=true` remain\ndetect-only when a category requires confirmation.\n\nThe script flags `--fix` and `--fix-all` are mutually exclusive. Exit\ncode is `0` if no ERRORs were produced, `1` otherwise. Every script run\n— regardless of outcome — writes exactly one `doctor.report` LogEntry\nvia the event-log MCP.\n\n### Default remediation workflow\n\nWhen a user invokes `\u002Fpk-doctor` or asks to check repository health, the\nagent must treat remediation as the default:\n\n1. Run `run_pk_doctor()` or the local `doctor.py` report.\n2. Build a queue of every `ERROR`, every `WARN`, and every finding whose\n   `action_required` is true, including actionable INFO findings.\n3. Resolve safe local fixes immediately through the owning script,\n   command, or MCP tool. Use the reported `fix_mcp_tool`,\n   `suggested_fix`, `action_kind`, and `default_agent_action` fields to\n   choose the narrowest correction.\n4. For migration findings, route through migration-management. Apply or\n   continue unambiguous migrations; reject malformed no-op migrations\n   only when the finding itself identifies the defect and a rejection\n   reason is clear. Ask the user only when the migration changes policy,\n   deletes data, has unclear intent, or conflicts with local work.\n5. For archive, policy, external-dependency, or user-confirmation\n   findings, either resolve them if the fix is already known and\n   non-destructive, create\u002Flink tracking when the finding calls for\n   tracking, or ask a concise question when a human decision is required.\n6. Rerun pk-doctor after remediation and report the final ERROR\u002FWARN and\n   action totals.\n\nIf the user explicitly says \"info only\", \"dry run\", \"report only\",\n\"check only\", \"no fixes\", or equivalent, stop after the report and do\nnot mutate anything. In that mode, still highlight the concrete next\nactions so the report is actionable.\n\n### The four Phase 1 checks\n\n1. **`schema_filename`** — walks every `context\u002F\u003Ckind>\u002F**\u002F*.md`. Loads\n   the matching schema from `src\u002Fcontext\u002Fschemas\u002F\u003Ckind>.yaml` (the LIVE\n   source, never the template mirror) and validates the file's frontmatter\n   against it. Also checks that the filename stem equals `metadata.id`\n   and that any date encoded in the filename matches `metadata.created`.\n   Phase 1 is WARN-only for rename suggestions — no auto-fix.\n2. **`sharding`** — logs must live under `context\u002Flogs\u002FYYYY\u002FMM\u002F` subdirs;\n   migrations must live under their `spec.state` subdir (`pending\u002F`,\n   `in-progress\u002F`, `applied\u002F`). WARN on any file in the wrong bucket.\n3. **`migrations`** — calls\n   `mcp__processkit-migration-management__list_migrations(state=\"pending\")`.\n   INFO for the pending count; WARN for any pending entry older than\n   14 days. Under `--fix=migrations` or `--fix-all`, interactively\n   prompts `Apply MIG-xxx? [y\u002FN\u002Fs]` per entry and invokes\n   `apply_migration` on yes.\n4. **`drift`** — subprocess-invokes `scripts\u002Fcheck-src-context-drift.sh`\n   from the repo root. Exit 0 → one INFO \"trees in sync\". Exit 1 →\n   one WARN per offending line reported by the script.\n5. **`commands_consistency`** — walks every\n   `context\u002Fskills\u002F*\u002F*\u002FSKILL.md` and verifies that each entry in\n   `metadata.processkit.commands:` has a matching `commands\u002F\u003Cname>.md`\n   file alongside the SKILL.md, uses the reserved `pk-` prefix, and has\n   matching `argument-hint` frontmatter. When `.claude\u002Fcommands\u002F` or\n   `.agents\u002Fskills\u002F` exists, it also verifies those harness projections\n   exactly match the canonical command set.\n\nAdditional checks added after Phase 1:\n\n- **`team_consistency`** — wraps `team-manager.check_all()`.\n- **`team_member_exports`** — reconciles active TeamMembers under `context\u002Fteam-members\u002F` with Claude sub-agent export files under `.claude\u002Fagents\u002F\u003Cslug>.md` (the path emitted by `team-manager.export_claude_subagent`). Emits WARN per active TeamMember missing an export (harness can't dispatch as `subagent_type`) and WARN per stale export whose slug no longer maps to an active TeamMember; INFO when the roster and the export dir are in sync. Detect-only — fix is to re-run `export_claude_subagent` (or `export_claude_subagents`) or delete the stale file. Lands the harness signal half of the sub-agent-dispatch clause from BACK-20260509_1317-WildPanda.\n- **`release_integrity`** — see DEC-20260422_1348-SnowyWolf-local-only-release-bulletproofing.\n- **`mcp_config_drift`** — reads `context\u002F.processkit-mcp-manifest.json`\n  (produced at release time by `scripts\u002Fgenerate-mcp-manifest.py`) and\n  recomputes the per-skill `mcp-config.json` sha256es. A missing manifest\n  or stale aggregate surfaces as WARN (run the generator). In a\n  derived-project context (`aibox.lock` + `.mcp.json` both present at\n  the repo root), any processkit server missing from `.mcp.json`'s\n  `mcpServers` map surfaces as ERROR with an explicit host-action hint\n  asking the owner to re-run the project installer\u002Fsync tool outside the\n  container. Exists because some host installers gate `.mcp.json`\n  re-merge on processkit version delta only, so per-skill MCP-config\n  edits within a release cycle may not reach derived projects until the\n  next version bump — see DEC-20260423_2049-VastLake.\n- **`server_header_drift`** — walks every `context\u002Fskills\u002Fprocesskit\u002F*\u002Fmcp\u002Fserver.py`,\n  hashes its PEP 723 inline metadata block (`# \u002F\u002F\u002F script` ... `# \u002F\u002F\u002F`),\n  and compares against the manifest's `per_server_header` baseline. WARN\n  on any drift listing the affected skill slugs — signal that someone\n  edited a dep header without regenerating the manifest, and the harness\n  needs a restart to pick up the new deps. Detect-only; the fix is\n  user-driven (regenerate manifest + restart harness). Per\n  DEC-20260424_0127-QuickPine (SharpBrook split, RapidSwan ships the\n  dep-drift half).\n- **`preauth_applied`** — compares\n  `context\u002Fskills\u002Fprocesskit\u002Fskill-gate\u002Fassets\u002Fpreauth.json` against\n  `.claude\u002Fsettings.json` and `.codex\u002Fconfig.toml`. WARNs when aibox has\n  not merged processkit MCP preauthorization into either harness config,\n  which is the common cause of reauthorization prompts after container\n  rebuilds.\n- **`context_consumption`** — emits an INFO-only estimate of processkit\n  context footprint by group (`startup`, command adapters, skill docs,\n  MCP configs) using `ceil(utf8 bytes \u002F 4)` as the provider-neutral\n  token heuristic. It is intentionally not an MCP tool, so measurement\n  does not add another tool schema to the harness context.\n  It also exposes a local checkpoint\u002Freport CLI:\n  ```sh\n  python context\u002Fskills\u002Fprocesskit\u002Fpk-doctor\u002Fscripts\u002Fchecks\u002Fcontext_consumption.py checkpoint \u003Clabel>\n  python context\u002Fskills\u002Fprocesskit\u002Fpk-doctor\u002Fscripts\u002Fchecks\u002Fcontext_consumption.py report \u003Cbefore> \u003Cafter>\n  ```\n  Checkpoints are JSON files under\n  `context\u002F.state\u002Fcontext-consumption\u002Fcheckpoints\u002F`. Reports compare\n  observed processkit payloads and label token counts as local estimates,\n  not provider-billed usage.\n- **`v2_contracts`** — validates v2 API contract invariants across WorkItem, Binding, Artifact, and LogEntry entities. Catches process-instance definitions missing their definition reference, time-window bindings without a recurrence rule, orphaned cost-policy artifacts (not bound to a budget), policy supersession chains with breaks, uncalibrated eval-spec judges, and stale or missing agent-card projections. Emits ERROR for all contract violations; detect-only.\n- **`context_hygiene`** — validates artifact naming policies (model-spec and model-profile use timestamped `ART-YYYYMMDD_HHMM-*` scheme), model binding integrity (role\u002FTeamMember defaults must target provider-neutral model-profile artifacts unless marked direct_model_pin), and cross-reference health. Also detects demoted schema kinds still present in src\u002F, detects archive candidates, warns on mixed binding filename styles, and checks sqlite-vec semantic index health. Emits WARN \u002F ERROR depending on severity; detect-only.\n- **`sensitive_data` allowlist** — intentional public or synthetic email\n  addresses may be listed in `.pk-doctor-allowlist.toml`:\n  ```toml\n  [sensitive_data.email_allowlist]\n  addresses = [\"info@projectious.work\", \"deploy@example.local\"]\n  ```\n  The allowlist applies only to email-address findings; secrets and other PII\n  are never suppressed.\n- **`runtime_health`** — owns in-container runtime probes that host-side\n  installer diagnostics cannot verify directly. Checks lnav availability\n  for the Prefix L structured log viewer, sqlite-vec import\u002Fload health\n  in the pk-doctor\u002FMCP runtime, Codex bubblewrap sandbox smoke status\n  when Codex is enabled, PID 1 sleep-infinity hygiene, cgroup\n  memory\u002Fpid\u002FOOM pressure, processkit Python MCP process counts,\n  PowerKit image\u002Fplugin tree, status plugin scripts, local render\n  helpers, and runtime-home write probes. Emits WARN only for\n  container-local actionable failures; host-only runs emit a skip INFO.\n- **`entity_storage_hygiene`** — validates local `context\u002F`\n  storage policy separately from template freshness: unmanaged host\n  artifacts, demoted legacy roots such as `context\u002Fmodels\u002F`,\n  root-level migration briefings, mixed root\u002Fsharded entity layouts,\n  mixed filename policies, placeholder `0000` timestamps, and\n  TeamMember private\u002Fslug policy explanations. Emits WARN for\n  actionable storage drift; legacy layouts and mixed filename policies\n  are migration work. `storage.root-migration-briefings` can be\n  auto-resolved via `--fix=entity_storage_hygiene` by archiving\n  non-completed CLI migration briefings.\n- **`schema_vocabulary`** — validates closed-vocabulary subtype fields across v2 entities by comparing frontmatter values against Schema-declared known-kinds\u002Fknown-types (Artifact kind, Binding type, WorkItem type, LogEntry event_type, Migration kind). `legacy_known_*` schema fields are not accepted as a terminal state; legacy values require an explicit data-fix Migration or a schema migration. Also validates that Migration entities include required v2 version metadata fields. Emits ERROR for unknown vocabulary values or missing version fields; detect-only.\n- **`migration_integrity`** — flags malformed pending migrations that exhibit the \"empty-baseline\" defect (same-version migrations with content in affected_groups\u002Faffected_files despite no-op intent, or affected_groups populated but affected_files empty). Per BACK-20260425_1711-CleverRiver, both patterns are defect signatures. Emits WARN with suggested fix to reject via migration-management MCP and refile the upstream bug; detect-only.\n- **`mcp_gateway`** — reports processkit-gateway MCP config health and harness registration state. Validates gateway config presence\u002Fstructure at `context\u002Fskills\u002Fprocesskit\u002Fprocesskit-gateway\u002Fmcp\u002Fmcp-config.json`, checks env vars and command launch target, detects mixed gateway+granular server registrations in `.mcp.json`, and warns on insecure daemon bindings. Emits INFO on healthy state, WARN on minor gaps (missing proxy --url, nonlocal daemon bind), ERROR on parsing failure; detect-only.\n- **`skill_dag`** — builds the skill-dependency directed-acyclic-graph (DAG) from `metadata.processkit.uses[*].skill` entries in every `context\u002Fskills\u002F**\u002FSKILL.md` and validates references exist, graph is acyclic, and layer constraints are honored (skill layer N only uses skills with layer ≤ N). Emits ERROR for missing references, cycles (with full path), or layer violations; detect-only.\n- **`v1_entity_drift`** — walks every registered v2 entity directory under `context\u002F` and surfaces files whose frontmatter still declares `apiVersion: …\u002Fv1`. Emits WARN per file with a hardcoded successor table (Actor→TeamMember, Process→Scope+Gate, StateMachine→lifecycle metadata, Model→Artifact(model-spec)); v1 files in append-only buckets (`context\u002Flogs\u002F`, `context\u002Fmigrations\u002Fapplied\u002F`) are downgraded to a single INFO since they are intentionally historical. Under `--fix=v1_entity_drift` (or `--fix-all`) the check interactively records `apply_migration` intents when a pending v1->v2 Migration already covers the target, otherwise records `propose_migration` intents — it never auto-creates migrations or hand-edits files. Per BACK-20260509_1318-KindSpruce.\n- **`agents_md_hygiene`** — validates the derived-project startup\n  policy surface. Checks root `AGENTS.md` for `pk-managed:*` blocks\n  (`pk-compliance-contract-v2`, `pk-commands`), compares managed block\n  content against `src\u002FAGENTS.md` or the latest template when available,\n  verifies processkit references for session start, skill routing,\n  entity IO, migrations, decision capture, sub-agent dispatch,\n  TeamMember\u002Fmodel binding, MCP topology, provider pointer files, and\n  command blocks, and flags stale guidance such as legacy actor roots,\n  `spec.x_aibox`, model-tier role tables, or duplicated provider-specific\n  policy. WARN findings include a `briefing` payload for a project agent\n  to reconcile the file after processkit upgrades. Detect-only; managed\n  block replacement and local policy merges remain user-reviewed.\n- **`sensitive_data`** — scans tracked text\u002Fconfig files for\n  deterministic high-signal secret and personal-data patterns: private\n  key blocks, provider tokens, JWTs, URL credentials, assigned\n  high-entropy secret values, email addresses, phone numbers, SSN-like\n  values, and credit-card-like numbers with a valid Luhn checksum.\n  Deterministic secret findings are ERROR\u002FWARN and require human\n  confirmation because the correct response may be rotation, redaction,\n  migration, or explicit policy acceptance. The check also emits one\n  INFO `sensitive-data.probabilistic-briefing` that tells the\n  derived-project agent what regex cannot prove: real names and aliases,\n  addresses, dates of birth, customer\u002Faccount IDs, medical\u002Ffinancial\u002F\n  employment data, screenshots, archives, database dumps, logs, and\n  non-standard short secrets. The briefing includes non-triggering\n  deterministic and probabilistic example prompts so an agent knows what\n  to search for without adding realistic secrets to the shipped scanner.\n  This is an advisory layer, not a complete DLP engine.\n- **`supply_chain`** — performs offline supply-chain hygiene checks for\n  dependency manifests\u002Flockfiles, license policy, and local scanner\n  outcomes:\n  - emits `ERROR` on denied licenses, missing application lockfiles, and\n    high\u002Fcritical vulnerabilities;\n  - emits `WARN` on unknown\u002Freview\u002Fallowed-policy misses and skipped\n    security scanners;\n  - emits `INFO` for inventory counts and SBOM discoveries.\n  Security\u002Foutdated\u002Fsupplier-quality checks stay advisory and require\n  explicit core-surface opt-in data; pk-doctor performs no outbound\n  network calls during this check.\n\n### What doctor will NEVER do\n\n- Hand-edit any file under `context\u002F`. All writes route through the\n  appropriate MCP tool (e.g. `apply_migration`).\n- Touch `context\u002Ftemplates\u002F` (read-only diff baseline).\n- Touch `.mcp.json` (merged file; use the generated per-skill\n  `mcp\u002Fmcp-config.json` and gateway catalog instead).\n- Write to `context\u002Flogs\u002F` directly — reports go through\n  `mcp__processkit-event-log__log_event`.\n- Treat a clean severity tally as done while actionable INFO findings\n  remain unresolved.\n- Stop after the report unless the user explicitly asked for info-only,\n  dry-run, report-only, check-only, or no-fix behavior.\n\n### Report shape\n\nAfter the stdout human-readable summary, the skill emits a single\n`doctor.report` LogEntry:\n\n```json\n{\n  \"event_type\": \"doctor.report\",\n  \"summary\": \"\u002Fpk-doctor --fix=migrations — 0 ERROR \u002F 2 WARN \u002F 3 INFO\",\n  \"details\": {\n    \"doctor_version\": \"1.0.0\",\n    \"invocation\": \"\u002Fpk-doctor --fix=migrations\",\n    \"categories\": {\n      \"schema_filename\": {\"ERROR\": 0, \"WARN\": 1, \"INFO\": 140},\n      \"sharding\": {\"ERROR\": 0, \"WARN\": 0, \"INFO\": 1},\n      \"migrations\": {\"ERROR\": 0, \"WARN\": 1, \"INFO\": 1},\n      \"drift\": {\"ERROR\": 0, \"WARN\": 0, \"INFO\": 1}\n    },\n    \"action_totals\": {\n      \"actionable\": 2,\n      \"needs_user_confirmation\": 1,\n      \"needs_tracking\": 0,\n      \"safe_fix\": 1,\n      \"migration_needed\": 1,\n      \"archive_needed\": 0,\n      \"policy_decision_needed\": 0,\n      \"external_dependency\": 0\n    },\n    \"top_findings\": [\n      {\n        \"severity\": \"WARN\",\n        \"id\": \"migration.stale-pending\",\n        \"entity_ref\": \"MIG-...\",\n        \"message\": \"...\",\n        \"action_required\": true,\n        \"action_kind\": \"migration_needed\",\n        \"default_agent_action\": \"create_migration\",\n        \"requires_user_confirmation\": true,\n        \"acceptable_resolution\": \"migrated\"\n      }\n    ],\n    \"fixes_applied\": [],\n    \"duration_ms\": 2340\n  }\n}\n```\n\n`top_findings` caps at 20 to keep the log entry compact; the full\nstdout report remains authoritative for the details of a specific run.\nThe structured JSON payload returned by `run_pk_doctor` includes the\nsame `action_totals` and actionability fields on every finding.\n\n### Interactive prompts\n\nInteractive prompts gate on `sys.stdin.isatty()`. If a `--fix` was\nrequested but the process isn't attached to a terminal, the skill emits\na `WARN` (`fix.non-interactive`) and skips the fix — it does not silently\nprompt and hang. `--yes` auto-confirms every prompt except any marked\n`data_loss=True` (none in Phase 1, but the hook is present).\n\n### `--since` scope\n\n`--since=\u003Cgit-ref>` restricts only the file-walk checks\n(`schema_filename`, `sharding`). The `migrations` and `drift` checks\nalways run full-scan: migrations pull from an MCP tool with its own\nstate, and drift compares entire trees.\n\n### Resolved policy defaults\n\nThese were agreed with the owner during shape review:\n\n- **YAML-datetime vs JSON-Schema string coercion** failures surface as\n  `WARN schema.datetime-coercion` (not ERROR). Parser-layer quirk.\n- **Command projections** — canonical command adapters live under\n  `context\u002Fskills\u002F**\u002Fcommands\u002F`; slash-capable harness projections live\n  under `.claude\u002Fcommands\u002F`; natural-language command shims live under\n  `.agents\u002Fskills\u002F`. `src\u002Fcontext\u002F` mirrors only the canonical source.\n- **Filename auto-rename** — WARN only in Phase 1. Phase 2 adds\n  orchestrated rename + reindex.\n\n## Gotchas\n\nAgent-specific failure modes — provider-neutral pause-and-self-check items:\n\n- **Confusing remediation mode with blind `--fix-all`.** The skill is\n  remediation-first by default, but that does not mean `doctor.py\n  --fix-all --yes` is safe. First run the report, read actionability,\n  then apply the narrowest known fix through the owning tool. Ask only\n  when a finding is destructive, policy-sensitive, external, or unclear.\n- **Stopping after a report when the user did not ask for info-only.**\n  `\u002Fpk-doctor` means \"check and clean up what can be cleaned up.\" If the\n  user only wants a briefing, they must say info-only, dry-run,\n  report-only, check-only, or no fixes.\n- **Treating drift warnings as noise.** The drift check exists because\n  v0.15.0–v0.18.0 shipped four releases with silent drift. If the\n  drift check produces a WARN, it has found a real divergence —\n  do not dismiss without investigating.\n- **Hand-editing files to \"clear\" a doctor warning.** Doctor warnings\n  are diagnostic. Fixing the underlying cause through the right MCP\n  write tool (not by editing the offending file directly) preserves\n  the audit trail. A WARN that reappears next run is more useful than\n  a hand-fixed tree with no record of the fix.\n- **Validating against `context\u002Ftemplates\u002F` schemas.** The template\n  mirror is a diff baseline, not the live schema source. Validation\n  must read `src\u002Fcontext\u002Fschemas\u002F\u003Ckind>.yaml` — the authoritative\n  copy. Validating against templates will produce stale false\n  positives after any schema evolution.\n- **Forgetting the `doctor.report` LogEntry on dry-run.** Even a\n  detect-only run must emit the report: it is the audit record of the\n  invocation. A run that prints to stdout but logs nothing strands\n  the next session with no record of what was checked.\n- **Running doctor mid-migration.** Doctor's `migrations` check will\n  list pending migrations and (under `--fix`) offer to apply them. If\n  you are already in the middle of a manual migration review, running\n  `--fix=migrations` can reorder work silently. Finish or pause the\n  manual review first.\n\n## Full reference\n\n### CLI contract\n\n```\ndoctor.py [--category=LIST] [--fix=LIST | --fix-all] [--since=REF] [--yes]\n```\n\n| Flag           | Meaning |\n|----------------|---------|\n| `--category`   | Comma list of categories to run. Default: all four. |\n| `--fix`        | Comma list of categories to enable fixes for. Mutex with `--fix-all`. |\n| `--fix-all`    | Enable fixes for every category that supports them. Mutex with `--fix`. |\n| `--since`      | Git ref; restricts `schema_filename` + `sharding` to files changed since. |\n| `--yes`        | Auto-confirm all non-`data_loss` fix prompts. |\n\n### Adding a check in Phase 2\n\n1. Drop a new module in `scripts\u002Fchecks\u002F\u003Cname>.py` exporting a\n   `run(ctx) -> list[CheckResult]` function (and optional\n   `run_fix(ctx, results) -> list[dict]`).\n2. Register it in `scripts\u002Fchecks\u002F__init__.py` in `REGISTRY`.\n3. Add a gotcha for that check here in SKILL.md.\n4. Mirror both files into `src\u002Fcontext\u002Fskills\u002Fprocesskit\u002Fpk-doctor\u002F`.\n\nThe aggregator in `doctor.py` does not need to change.\n\n### What this skill does NOT do (yet)\n\n- Validate cross-references (Phase 2).\n- Check `PROVENANCE.toml` drift (Phase 2).\n- Check `mcp-config.json` integrity (Phase 2).\n- Smoke-import MCP servers (Phase 2).\n- Detect version skew between `aibox.lock` and installed SKILL.md pins\n  (Phase 2).\n- Expose itself as an MCP server for machine invocation (Phase 3, only\n  if demand shows up).\n",{"name":5,"metadata":54,"description":79},{"processkit":55},{"id":56,"uses":57,"layer":67,"created":68,"version":69,"category":70,"commands":71,"provides":75,"apiVersion":78},"SKILL-pk-doctor",[58,61,64],{"skill":59,"purpose":60},"event-log","Emit the single `doctor.report` LogEntry per run (the audit trail for every invocation).",{"skill":62,"purpose":63},"migration-management","List pending migrations for the `migrations` check and apply them under `--fix`.",{"skill":65,"purpose":66},"index-management","(future) surface index errors alongside doctor findings in Phase 2.",3,1776643200000,"1.0.0","processkit",[72],{"args":73,"name":5,"description":74},"[info-only | dry-run | --category=... | --fix=... | --fix-all | --since=\u003Cref> | --yes]","Run health checks and resolve actionable findings by default.",{"mcp_tools":76},[77],"run_pk_doctor","processkit.projectious.work\u002Fv2","Aggregator health-check for a processkit-managed repository. Runs a fixed suite of checks over the live context\u002F tree plus src\u002Fcontext\u002F drift, then resolves ERROR \u002F WARN \u002F actionable INFO findings by default. Use when the user invokes `\u002Fpk-doctor`, asks \"is this repo healthy\", or has just finished an upgrade and wants a sanity pass. Info-only mode is opt-in: stop after the report only when the user explicitly says info-only, dry-run, report-only, check-only, or no fixes.\n","main","4876a92620ee792aac2ac9a1a806322d8a572367","archive\u002Fprocesskit-install-20261101\u002Fcontext\u002Fskills\u002Fprocesskit\u002Fpk-doctor","https:\u002F\u002Fgithub.com\u002Fprojectious-work\u002Fprocesskit","https:\u002F\u002Fprojectious-work.github.io\u002Fprocesskit\u002F","https:\u002F\u002Fgithub.com\u002Fprojectious-work","2026-11-11 23:24:18",{"source":8,"skillName":5,"commands":88,"prompt":98,"zipAvailable":99},[89,92,95],{"tool":90,"command":91},"npx","npx skills add projectious-work\u002Fprocesskit --skill pk-doctor",{"tool":93,"command":94},"bunx","bunx skills add projectious-work\u002Fprocesskit --skill pk-doctor",{"tool":96,"command":97},"pnpm","pnpm dlx skills add projectious-work\u002Fprocesskit --skill pk-doctor","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fprojectious-work\u002Fprocesskit\u002Fblob\u002Fmain\u002Farchive\u002Fprocesskit-install-20261101\u002Fcontext\u002Fskills\u002Fprocesskit\u002Fpk-doctor\u002FSKILL.md",true,[101,107,111,117,121,125,129,133,137,141,145,149,153,157,161,165,169,173,177,181,185,189,193,197,201,205,209,213,217,221,225,229,233,236,240,244,248],{"path":102,"resourceType":103,"resourceTypeDesc":104,"sizeBytes":105,"sourceHtmlUrl":106},"commands\u002Fpk-doctor.md","REFERENCE","参考文档",801,"https:\u002F\u002Fgithub.com\u002Fprojectious-work\u002Fprocesskit\u002Fblob\u002Fmain\u002Farchive\u002Fprocesskit-install-20261101\u002Fcontext\u002Fskills\u002Fprocesskit\u002Fpk-doctor\u002Fcommands\u002Fpk-doctor.md",{"path":108,"resourceType":103,"resourceTypeDesc":104,"sizeBytes":109,"sourceHtmlUrl":110},"mcp\u002Fmcp-config.json",201,"https:\u002F\u002Fgithub.com\u002Fprojectious-work\u002Fprocesskit\u002Fblob\u002Fmain\u002Farchive\u002Fprocesskit-install-20261101\u002Fcontext\u002Fskills\u002Fprocesskit\u002Fpk-doctor\u002Fmcp\u002Fmcp-config.json",{"path":112,"resourceType":113,"resourceTypeDesc":114,"sizeBytes":115,"sourceHtmlUrl":116},"mcp\u002Fserver.py","SCRIPT","脚本",4545,"https:\u002F\u002Fgithub.com\u002Fprojectious-work\u002Fprocesskit\u002Fblob\u002Fmain\u002Farchive\u002Fprocesskit-install-20261101\u002Fcontext\u002Fskills\u002Fprocesskit\u002Fpk-doctor\u002Fmcp\u002Fserver.py",{"path":118,"resourceType":113,"resourceTypeDesc":114,"sizeBytes":119,"sourceHtmlUrl":120},"scripts\u002Fchecks\u002F__init__.py",2243,"https:\u002F\u002Fgithub.com\u002Fprojectious-work\u002Fprocesskit\u002Fblob\u002Fmain\u002Farchive\u002Fprocesskit-install-20261101\u002Fcontext\u002Fskills\u002Fprocesskit\u002Fpk-doctor\u002Fscripts\u002Fchecks\u002F__init__.py",{"path":122,"resourceType":113,"resourceTypeDesc":114,"sizeBytes":123,"sourceHtmlUrl":124},"scripts\u002Fchecks\u002Fagents_md_hygiene.py",17566,"https:\u002F\u002Fgithub.com\u002Fprojectious-work\u002Fprocesskit\u002Fblob\u002Fmain\u002Farchive\u002Fprocesskit-install-20261101\u002Fcontext\u002Fskills\u002Fprocesskit\u002Fpk-doctor\u002Fscripts\u002Fchecks\u002Fagents_md_hygiene.py",{"path":126,"resourceType":113,"resourceTypeDesc":114,"sizeBytes":127,"sourceHtmlUrl":128},"scripts\u002Fchecks\u002Fcommands_consistency.py",9198,"https:\u002F\u002Fgithub.com\u002Fprojectious-work\u002Fprocesskit\u002Fblob\u002Fmain\u002Farchive\u002Fprocesskit-install-20261101\u002Fcontext\u002Fskills\u002Fprocesskit\u002Fpk-doctor\u002Fscripts\u002Fchecks\u002Fcommands_consistency.py",{"path":130,"resourceType":113,"resourceTypeDesc":114,"sizeBytes":131,"sourceHtmlUrl":132},"scripts\u002Fchecks\u002Fcommon.py",5941,"https:\u002F\u002Fgithub.com\u002Fprojectious-work\u002Fprocesskit\u002Fblob\u002Fmain\u002Farchive\u002Fprocesskit-install-20261101\u002Fcontext\u002Fskills\u002Fprocesskit\u002Fpk-doctor\u002Fscripts\u002Fchecks\u002Fcommon.py",{"path":134,"resourceType":113,"resourceTypeDesc":114,"sizeBytes":135,"sourceHtmlUrl":136},"scripts\u002Fchecks\u002Fcontext_consumption.py",12448,"https:\u002F\u002Fgithub.com\u002Fprojectious-work\u002Fprocesskit\u002Fblob\u002Fmain\u002Farchive\u002Fprocesskit-install-20261101\u002Fcontext\u002Fskills\u002Fprocesskit\u002Fpk-doctor\u002Fscripts\u002Fchecks\u002Fcontext_consumption.py",{"path":138,"resourceType":113,"resourceTypeDesc":114,"sizeBytes":139,"sourceHtmlUrl":140},"scripts\u002Fchecks\u002Fcontext_hygiene.py",24062,"https:\u002F\u002Fgithub.com\u002Fprojectious-work\u002Fprocesskit\u002Fblob\u002Fmain\u002Farchive\u002Fprocesskit-install-20261101\u002Fcontext\u002Fskills\u002Fprocesskit\u002Fpk-doctor\u002Fscripts\u002Fchecks\u002Fcontext_hygiene.py",{"path":142,"resourceType":113,"resourceTypeDesc":114,"sizeBytes":143,"sourceHtmlUrl":144},"scripts\u002Fchecks\u002Fdoctor_boundary.py",2824,"https:\u002F\u002Fgithub.com\u002Fprojectious-work\u002Fprocesskit\u002Fblob\u002Fmain\u002Farchive\u002Fprocesskit-install-20261101\u002Fcontext\u002Fskills\u002Fprocesskit\u002Fpk-doctor\u002Fscripts\u002Fchecks\u002Fdoctor_boundary.py",{"path":146,"resourceType":113,"resourceTypeDesc":114,"sizeBytes":147,"sourceHtmlUrl":148},"scripts\u002Fchecks\u002Fdrift.py",2961,"https:\u002F\u002Fgithub.com\u002Fprojectious-work\u002Fprocesskit\u002Fblob\u002Fmain\u002Farchive\u002Fprocesskit-install-20261101\u002Fcontext\u002Fskills\u002Fprocesskit\u002Fpk-doctor\u002Fscripts\u002Fchecks\u002Fdrift.py",{"path":150,"resourceType":113,"resourceTypeDesc":114,"sizeBytes":151,"sourceHtmlUrl":152},"scripts\u002Fchecks\u002Fentity_storage_hygiene.py",21367,"https:\u002F\u002Fgithub.com\u002Fprojectious-work\u002Fprocesskit\u002Fblob\u002Fmain\u002Farchive\u002Fprocesskit-install-20261101\u002Fcontext\u002Fskills\u002Fprocesskit\u002Fpk-doctor\u002Fscripts\u002Fchecks\u002Fentity_storage_hygiene.py",{"path":154,"resourceType":113,"resourceTypeDesc":114,"sizeBytes":155,"sourceHtmlUrl":156},"scripts\u002Fchecks\u002Fid_vocabulary.py",8212,"https:\u002F\u002Fgithub.com\u002Fprojectious-work\u002Fprocesskit\u002Fblob\u002Fmain\u002Farchive\u002Fprocesskit-install-20261101\u002Fcontext\u002Fskills\u002Fprocesskit\u002Fpk-doctor\u002Fscripts\u002Fchecks\u002Fid_vocabulary.py",{"path":158,"resourceType":113,"resourceTypeDesc":114,"sizeBytes":159,"sourceHtmlUrl":160},"scripts\u002Fchecks\u002Fmcp_config_drift.py",10945,"https:\u002F\u002Fgithub.com\u002Fprojectious-work\u002Fprocesskit\u002Fblob\u002Fmain\u002Farchive\u002Fprocesskit-install-20261101\u002Fcontext\u002Fskills\u002Fprocesskit\u002Fpk-doctor\u002Fscripts\u002Fchecks\u002Fmcp_config_drift.py",{"path":162,"resourceType":113,"resourceTypeDesc":114,"sizeBytes":163,"sourceHtmlUrl":164},"scripts\u002Fchecks\u002Fmcp_gateway.py",6616,"https:\u002F\u002Fgithub.com\u002Fprojectious-work\u002Fprocesskit\u002Fblob\u002Fmain\u002Farchive\u002Fprocesskit-install-20261101\u002Fcontext\u002Fskills\u002Fprocesskit\u002Fpk-doctor\u002Fscripts\u002Fchecks\u002Fmcp_gateway.py",{"path":166,"resourceType":113,"resourceTypeDesc":114,"sizeBytes":167,"sourceHtmlUrl":168},"scripts\u002Fchecks\u002Fmigration_integrity.py",4236,"https:\u002F\u002Fgithub.com\u002Fprojectious-work\u002Fprocesskit\u002Fblob\u002Fmain\u002Farchive\u002Fprocesskit-install-20261101\u002Fcontext\u002Fskills\u002Fprocesskit\u002Fpk-doctor\u002Fscripts\u002Fchecks\u002Fmigration_integrity.py",{"path":170,"resourceType":113,"resourceTypeDesc":114,"sizeBytes":171,"sourceHtmlUrl":172},"scripts\u002Fchecks\u002Fmigrations.py",7927,"https:\u002F\u002Fgithub.com\u002Fprojectious-work\u002Fprocesskit\u002Fblob\u002Fmain\u002Farchive\u002Fprocesskit-install-20261101\u002Fcontext\u002Fskills\u002Fprocesskit\u002Fpk-doctor\u002Fscripts\u002Fchecks\u002Fmigrations.py",{"path":174,"resourceType":113,"resourceTypeDesc":114,"sizeBytes":175,"sourceHtmlUrl":176},"scripts\u002Fchecks\u002Fpreauth_applied.py",18617,"https:\u002F\u002Fgithub.com\u002Fprojectious-work\u002Fprocesskit\u002Fblob\u002Fmain\u002Farchive\u002Fprocesskit-install-20261101\u002Fcontext\u002Fskills\u002Fprocesskit\u002Fpk-doctor\u002Fscripts\u002Fchecks\u002Fpreauth_applied.py",{"path":178,"resourceType":113,"resourceTypeDesc":114,"sizeBytes":179,"sourceHtmlUrl":180},"scripts\u002Fchecks\u002Frelease_integrity.py",6453,"https:\u002F\u002Fgithub.com\u002Fprojectious-work\u002Fprocesskit\u002Fblob\u002Fmain\u002Farchive\u002Fprocesskit-install-20261101\u002Fcontext\u002Fskills\u002Fprocesskit\u002Fpk-doctor\u002Fscripts\u002Fchecks\u002Frelease_integrity.py",{"path":182,"resourceType":113,"resourceTypeDesc":114,"sizeBytes":183,"sourceHtmlUrl":184},"scripts\u002Fchecks\u002Fruntime_health.py",13936,"https:\u002F\u002Fgithub.com\u002Fprojectious-work\u002Fprocesskit\u002Fblob\u002Fmain\u002Farchive\u002Fprocesskit-install-20261101\u002Fcontext\u002Fskills\u002Fprocesskit\u002Fpk-doctor\u002Fscripts\u002Fchecks\u002Fruntime_health.py",{"path":186,"resourceType":113,"resourceTypeDesc":114,"sizeBytes":187,"sourceHtmlUrl":188},"scripts\u002Fchecks\u002Fschema_filename.py",22994,"https:\u002F\u002Fgithub.com\u002Fprojectious-work\u002Fprocesskit\u002Fblob\u002Fmain\u002Farchive\u002Fprocesskit-install-20261101\u002Fcontext\u002Fskills\u002Fprocesskit\u002Fpk-doctor\u002Fscripts\u002Fchecks\u002Fschema_filename.py",{"path":190,"resourceType":113,"resourceTypeDesc":114,"sizeBytes":191,"sourceHtmlUrl":192},"scripts\u002Fchecks\u002Fschema_vocabulary.py",7437,"https:\u002F\u002Fgithub.com\u002Fprojectious-work\u002Fprocesskit\u002Fblob\u002Fmain\u002Farchive\u002Fprocesskit-install-20261101\u002Fcontext\u002Fskills\u002Fprocesskit\u002Fpk-doctor\u002Fscripts\u002Fchecks\u002Fschema_vocabulary.py",{"path":194,"resourceType":113,"resourceTypeDesc":114,"sizeBytes":195,"sourceHtmlUrl":196},"scripts\u002Fchecks\u002Fsensitive_data.py",18930,"https:\u002F\u002Fgithub.com\u002Fprojectious-work\u002Fprocesskit\u002Fblob\u002Fmain\u002Farchive\u002Fprocesskit-install-20261101\u002Fcontext\u002Fskills\u002Fprocesskit\u002Fpk-doctor\u002Fscripts\u002Fchecks\u002Fsensitive_data.py",{"path":198,"resourceType":113,"resourceTypeDesc":114,"sizeBytes":199,"sourceHtmlUrl":200},"scripts\u002Fchecks\u002Fserver_header_drift.py",7910,"https:\u002F\u002Fgithub.com\u002Fprojectious-work\u002Fprocesskit\u002Fblob\u002Fmain\u002Farchive\u002Fprocesskit-install-20261101\u002Fcontext\u002Fskills\u002Fprocesskit\u002Fpk-doctor\u002Fscripts\u002Fchecks\u002Fserver_header_drift.py",{"path":202,"resourceType":113,"resourceTypeDesc":114,"sizeBytes":203,"sourceHtmlUrl":204},"scripts\u002Fchecks\u002Fsharding.py",11457,"https:\u002F\u002Fgithub.com\u002Fprojectious-work\u002Fprocesskit\u002Fblob\u002Fmain\u002Farchive\u002Fprocesskit-install-20261101\u002Fcontext\u002Fskills\u002Fprocesskit\u002Fpk-doctor\u002Fscripts\u002Fchecks\u002Fsharding.py",{"path":206,"resourceType":113,"resourceTypeDesc":114,"sizeBytes":207,"sourceHtmlUrl":208},"scripts\u002Fchecks\u002Fskill_dag.py",9392,"https:\u002F\u002Fgithub.com\u002Fprojectious-work\u002Fprocesskit\u002Fblob\u002Fmain\u002Farchive\u002Fprocesskit-install-20261101\u002Fcontext\u002Fskills\u002Fprocesskit\u002Fpk-doctor\u002Fscripts\u002Fchecks\u002Fskill_dag.py",{"path":210,"resourceType":113,"resourceTypeDesc":114,"sizeBytes":211,"sourceHtmlUrl":212},"scripts\u002Fchecks\u002Fsupply_chain.py",21847,"https:\u002F\u002Fgithub.com\u002Fprojectious-work\u002Fprocesskit\u002Fblob\u002Fmain\u002Farchive\u002Fprocesskit-install-20261101\u002Fcontext\u002Fskills\u002Fprocesskit\u002Fpk-doctor\u002Fscripts\u002Fchecks\u002Fsupply_chain.py",{"path":214,"resourceType":113,"resourceTypeDesc":114,"sizeBytes":215,"sourceHtmlUrl":216},"scripts\u002Fchecks\u002Fteam_consistency.py",3951,"https:\u002F\u002Fgithub.com\u002Fprojectious-work\u002Fprocesskit\u002Fblob\u002Fmain\u002Farchive\u002Fprocesskit-install-20261101\u002Fcontext\u002Fskills\u002Fprocesskit\u002Fpk-doctor\u002Fscripts\u002Fchecks\u002Fteam_consistency.py",{"path":218,"resourceType":113,"resourceTypeDesc":114,"sizeBytes":219,"sourceHtmlUrl":220},"scripts\u002Fchecks\u002Fteam_member_exports.py",5249,"https:\u002F\u002Fgithub.com\u002Fprojectious-work\u002Fprocesskit\u002Fblob\u002Fmain\u002Farchive\u002Fprocesskit-install-20261101\u002Fcontext\u002Fskills\u002Fprocesskit\u002Fpk-doctor\u002Fscripts\u002Fchecks\u002Fteam_member_exports.py",{"path":222,"resourceType":113,"resourceTypeDesc":114,"sizeBytes":223,"sourceHtmlUrl":224},"scripts\u002Fchecks\u002Fv1_entity_drift.py",12618,"https:\u002F\u002Fgithub.com\u002Fprojectious-work\u002Fprocesskit\u002Fblob\u002Fmain\u002Farchive\u002Fprocesskit-install-20261101\u002Fcontext\u002Fskills\u002Fprocesskit\u002Fpk-doctor\u002Fscripts\u002Fchecks\u002Fv1_entity_drift.py",{"path":226,"resourceType":113,"resourceTypeDesc":114,"sizeBytes":227,"sourceHtmlUrl":228},"scripts\u002Fchecks\u002Fv2_contracts.py",14929,"https:\u002F\u002Fgithub.com\u002Fprojectious-work\u002Fprocesskit\u002Fblob\u002Fmain\u002Farchive\u002Fprocesskit-install-20261101\u002Fcontext\u002Fskills\u002Fprocesskit\u002Fpk-doctor\u002Fscripts\u002Fchecks\u002Fv2_contracts.py",{"path":230,"resourceType":113,"resourceTypeDesc":114,"sizeBytes":231,"sourceHtmlUrl":232},"scripts\u002Fdoctor.py",13421,"https:\u002F\u002Fgithub.com\u002Fprojectious-work\u002Fprocesskit\u002Fblob\u002Fmain\u002Farchive\u002Fprocesskit-install-20261101\u002Fcontext\u002Fskills\u002Fprocesskit\u002Fpk-doctor\u002Fscripts\u002Fdoctor.py",{"path":234,"resourceType":113,"resourceTypeDesc":114,"sizeBytes":14,"sourceHtmlUrl":235},"scripts\u002Ffixtures\u002F.gitkeep","https:\u002F\u002Fgithub.com\u002Fprojectious-work\u002Fprocesskit\u002Fblob\u002Fmain\u002Farchive\u002Fprocesskit-install-20261101\u002Fcontext\u002Fskills\u002Fprocesskit\u002Fpk-doctor\u002Fscripts\u002Ffixtures\u002F.gitkeep",{"path":237,"resourceType":113,"resourceTypeDesc":114,"sizeBytes":238,"sourceHtmlUrl":239},"scripts\u002Ftest_doctor.py",104911,"https:\u002F\u002Fgithub.com\u002Fprojectious-work\u002Fprocesskit\u002Fblob\u002Fmain\u002Farchive\u002Fprocesskit-install-20261101\u002Fcontext\u002Fskills\u002Fprocesskit\u002Fpk-doctor\u002Fscripts\u002Ftest_doctor.py",{"path":241,"resourceType":113,"resourceTypeDesc":114,"sizeBytes":242,"sourceHtmlUrl":243},"scripts\u002Ftest_id_vocabulary.py",1029,"https:\u002F\u002Fgithub.com\u002Fprojectious-work\u002Fprocesskit\u002Fblob\u002Fmain\u002Farchive\u002Fprocesskit-install-20261101\u002Fcontext\u002Fskills\u002Fprocesskit\u002Fpk-doctor\u002Fscripts\u002Ftest_id_vocabulary.py",{"path":245,"resourceType":113,"resourceTypeDesc":114,"sizeBytes":246,"sourceHtmlUrl":247},"scripts\u002Ftest_pk_doctor_json.py",6315,"https:\u002F\u002Fgithub.com\u002Fprojectious-work\u002Fprocesskit\u002Fblob\u002Fmain\u002Farchive\u002Fprocesskit-install-20261101\u002Fcontext\u002Fskills\u002Fprocesskit\u002Fpk-doctor\u002Fscripts\u002Ftest_pk_doctor_json.py",{"path":249,"resourceType":113,"resourceTypeDesc":114,"sizeBytes":250,"sourceHtmlUrl":251},"scripts\u002Ftest_pk_doctor_mcp.py",1669,"https:\u002F\u002Fgithub.com\u002Fprojectious-work\u002Fprocesskit\u002Fblob\u002Fmain\u002Farchive\u002Fprocesskit-install-20261101\u002Fcontext\u002Fskills\u002Fprocesskit\u002Fpk-doctor\u002Fscripts\u002Ftest_pk_doctor_mcp.py",[253,260,267,273,297,316],{"skillNo":254,"name":255,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":256,"categories":257,"keywords":258,"publishedAt":259},"01M7665G2ED3624R3X93304265","standard-ui-design","https:\u002F\u002Fgithub.com\u002Fprojectious-work\u002Fprocesskit\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Fstandard-ui-design\u002FSKILL.md",[],[],"2026-11-12 13:55:43",{"skillNo":261,"name":262,"nameZh":6,"description":263,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":264,"categories":265,"keywords":266,"publishedAt":259},"01M76651QTC6W0HJ4AFW2MMR3V","standard-agent-use","Company standard for working with AI agents. Use when deciding what to delegate to agents, capturing open work, keeping agent memory or knowledge notes, improving an agent workflow, or introducing agent work in a team or for a client.","https:\u002F\u002Fgithub.com\u002Fprojectious-work\u002Fprocesskit\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Fstandard-agent-use\u002FSKILL.md",[],[],{"skillNo":268,"name":269,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":270,"categories":271,"keywords":272,"publishedAt":259},"01M76650NN4T8P241021NBR64S","standard-agent-building","https:\u002F\u002Fgithub.com\u002Fprojectious-work\u002Fprocesskit\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Fstandard-agent-building\u002FSKILL.md",[],[],{"skillNo":274,"name":275,"nameZh":6,"description":276,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":277,"categories":278,"keywords":280,"publishedAt":49},"01M6X5MTC89B4ZGXW3JVSQCPF8","skill-management","Manage Skill package manifests as v1 ontology entities.","https:\u002F\u002Fgithub.com\u002Fprojectious-work\u002Fprocesskit\u002Fblob\u002Fmain\u002Fsrc\u002Fcontext\u002Fskills\u002Fprocesskit\u002Fskill-management\u002FSKILL.md",[279],{"categoryNo":19,"categoryName":20,"taxonomyType":21,"taxonomyTypeDesc":22},[281,282,283,284,285,286,287,288,289,290,291,292,275,293,294,295,296],"management","skill management","ontology","v1","create","as","work","transition","get","manage","update","alpha","list","package","skill package","manifests",{"skillNo":298,"name":299,"nameZh":6,"description":300,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":301,"categories":302,"keywords":304,"publishedAt":49},"01M6X5MCV74NHW5MEDNHZRF2C8","schema-management","Manage generated processkit schema contracts through deterministic MCP rebuild and inspection workflows.","https:\u002F\u002Fgithub.com\u002Fprojectious-work\u002Fprocesskit\u002Fblob\u002Fmain\u002Fsrc\u002Fcontext\u002Fskills\u002Fprocesskit\u002Fschema-management\u002FSKILL.md",[303],{"categoryNo":19,"categoryName":20,"taxonomyType":21,"taxonomyTypeDesc":22},[305,306,281,307,308,309,310,311,312,313,314,315,287,289,290,292,35],"schema","schema management","contract","rebuild","contracts","workflows","wed","generated","MCP","validation","deterministic",{"skillNo":317,"name":318,"nameZh":6,"description":319,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":320,"categories":321,"keywords":322,"publishedAt":49},"01M6X5M5G6YJ96AMZGSCZJ1D13","proposition-management","Record claims, risks, beliefs, world facts, WSJF estimates, and assumptions through the Proposition interface.","https:\u002F\u002Fgithub.com\u002Fprojectious-work\u002Fprocesskit\u002Fblob\u002Fmain\u002Fsrc\u002Fcontext\u002Fskills\u002Fprocesskit\u002Fproposition-management\u002FSKILL.md",[],[323,281,324,325,285,326,327,328,287,289,329,330,291,331,292,332,333,334,335,46],"proposition management","proposition","interface","claims","record","world","creation","event","WSJF","index","updates","query","facts"]