[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M6X5FX4PYGAN93423CQX6JRA":3,"skill-same:01M6X5FX4PYGAN93423CQX6JRA":49},{"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":18,"publishedAt":23,"linkOnly":24,"markdownContent":25,"markdownBody":26,"frontMatter":27,"refName":28,"commitSha":29,"skillDirectory":30,"forksCount":14,"repositoryHtmlUrl":31,"repositoryHomepageUrl":32,"ownerHtmlUrl":33,"ownerBlogUrl":6,"ownerDescription":6,"updatedAt":34,"install":35,"resources":48},"01M6X5FX4PYGAN93423CQX6JRA","pk-team-create",null,"Command: pk-team-create","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\u002Fharness\u002Fclaude-skills\u002Fpk-team-create\u002FSKILL.md",[],[19,20,21,22],"create","team","pk team create","command","2026-11-10 10:12:30",false,"---\nname: pk-team-create\ndescription: \"Command: pk-team-create\"\n---\n\n# Command: pk-team-create\n\nFull team derivation from scratch (v2 \u002F catalog-driven). Selects\ncatalog Roles via the archetype-catalog mapping, opens RoleSlots\nunder the chartering Scope, fills them with TeamMembers, rewrites\n`context\u002Fteam\u002Froster.md`, and emits a chartering DecisionRecord that\nsupersedes the prior team DecisionRecord. **No archetype Role\nentities are written.**\n\n## Syntax\n\n```\npk-team-create\n  --chartering-scope \u003CSCOPE-id>      # required — Scope that owns the team's RoleSlots\n  --subscription \u003Cprovider>:\u003Ctier>   # e.g. anthropic:max-5x\n  --providers \u003Clist|\"any\">           # comma-separated; \"any\" = all accessible\n  --parallelism-cap \u003Cint>            # max RoleSlots opened per archetype, default 5\n  --governance-floor \u003C0-5>           # G-score floor, default 3\n  [--weight-overrides \u003Cjson>]        # {\"C\":0.60,\"K\":0.20,\"L\":0.10,\"G\":0.10}\n                                     # CLI > DEC-*-TeamWeights > skill defaults\n  [--threshold-overrides \u003Cjson>]     # {\"heavy_min\":0.70,\"medium_min\":0.40}\n                                     # CLI > DEC-*-TeamWeights > skill defaults\n  [--landscape-artifact \u003CART-id>]    # explicit landscape override; skips discovery\n  [--landscape \u003Cartifact-id>]        # alias for --landscape-artifact (kept for compat)\n  [--archetype-catalog-mapping \u003Cpath>]\n                                     # CLI override of the archetype→catalog\n                                     # mapping (replace mode). Project override\n                                     # at context\u002Fteam\u002Farchetype-catalog-mapping.yaml\n                                     # is auto-detected; this flag wins over both.\n  [--budget-drift-threshold \u003Cfloat>] # drift alert threshold %, default 20\n  [--projection-method \u003Cmethod>]     # heuristic | model-recommender-quote | manual\n                                     # default: heuristic\n  [--dry-run]                        # print plan; write nothing\n```\n\n## Process (sequential)\n\n### Step 1 — Resolve landscape snapshot\n\nThree-level precedence (see `references\u002Flandscape-resolution.md`):\n\n1. If `--landscape-artifact \u003CART-id>` (or alias `--landscape`) is\n   supplied, load that artifact directly. Record\n   `inputs_snapshot.landscape_artifact_source: \"explicit\"`.\n2. Else query artifact index for the most-recently-created artifact\n   tagged both `landscape-summary` **and** `landscape-summary-project`\n   whose `spec.project_id` matches the current project context.\n   Record `inputs_snapshot.landscape_artifact_source: \"project-tag\"`.\n3. Else fall back to the most-recently-created artifact tagged\n   `landscape-summary` (kit default). Record\n   `inputs_snapshot.landscape_artifact_source: \"kit-default\"`.\n\nIf no artifact is found at any level, **abort** with:\n> \"No landscape-summary artifact found. Run the Haiku landscape-ingest\n> task to produce an ART-*-LandscapeSnapshot artifact before running\n> pk-team-create.\"\n\nIf the resolved artifact is older than 90 days, emit a warning but\ncontinue. Record the artifact ID and its `created` date in the\nDecisionRecord `inputs_snapshot`.\n\n### Step 2 — Query accessible models\n\n```\nmodel-recommender.check_availability()\nmodel-recommender.query_models(\n  providers=\u003C--providers>,\n  G_floor=\u003C--governance-floor>,\n  apply_user_filter=true\n)\nmodel-recommender.get_pricing(sort_by=\"value_score\")\n```\n\nFilter out any model with availability != \"operational\" unless the\nowner explicitly passes `--allow-degraded`.\n\n### Step 3 — Resolve weights + thresholds, then apply tiering formula\n\n**Weight resolution (before scoring):**\n\n```\nCLI --weight-overrides \u003Cjson>\n  > DEC-*-TeamWeights.weight_overrides (tag: team-weights-override,\n      most recent accepted; see references\u002Fteam-weights-decision-schema.md)\n    > skill defaults: {C:0.60, K:0.20, L:0.10, G:0.10}\n```\n\n**Threshold resolution (before classification):**\n\n```\nCLI --threshold-overrides \u003Cjson>\n  > DEC-*-TeamWeights.tier_thresholds (same DEC; validated on load)\n    > skill defaults: {heavy_min:0.70, medium_min:0.40}\n```\n\nValidate any threshold override against the rules in\n`references\u002Fteam-weights-decision-schema.md` (§Threshold validation\nrules). Violations are hard errors — abort before scoring.\n\nRecord `inputs_snapshot.weights_source` and\n`inputs_snapshot.tier_thresholds_source` with the resolution level\nused (`\"cli\"`, `\"dec-team-weights\"`, or `\"skill-default\"`), and\nstore the actual values in `inputs_snapshot.weights` and\n`inputs_snapshot.tier_thresholds`.\n\n**Scoring:**\n\nFor each accessible candidate model `m`:\n\n1. Compute raw dimension values:\n   - `C_raw(m)` = (swe_bench_verified + swe_bench_pro +\n                   community_rating) \u002F 3\n   - `K_raw(m)` = 1 \u002F (output_price_per_1M × quota_burn_vs_flagship)\n   - `L_raw(m)` = output_tokens_per_sec\n   - `G_raw(m)` = G-score from model-recommender (0–5 scale,\n                  normalise to 0–1 by dividing by 5)\n\n2. Min-max normalise each dimension across the full candidate set:\n   `norm(x) = clamp((x - min) \u002F (max - min), 0.01, 1.00)`\n\n3. Compute TierScore using resolved weights `{C, K, L, G}`:\n   `TierScore(m) = C×norm(C) + K×norm(K) + L×norm(L) + G×norm(G)`\n\n4. Classify using resolved thresholds:\n   - TierScore ≥ `heavy_min` → **heavy**\n   - TierScore ≥ `medium_min` and \u003C `heavy_min` → **medium**\n   - TierScore \u003C `medium_min` → **light**\n\n5. Tier-collapse rule: if fewer than 3 distinct tiers result, promote\n   the two highest-scoring light models to medium.\n\nSee `references\u002Ftiering-formula.md` for the full formula and worked\nexample.\n\n### Step 4 — Load archetype-catalog mapping + role-archetypes override, then map archetypes\n\n**Layer A — archetype-catalog mapping (eager validation — before archetype mapping):**\n\nResolve the archetype → catalog `(ROLE-id, seniority)` mapping with\nthree-level precedence:\n\n1. `--archetype-catalog-mapping \u003Cpath>` flag (CLI; replace semantics).\n2. `context\u002Fteam\u002Farchetype-catalog-mapping.yaml` (project; delta\n   semantics by default — top-level `override_semantics: replace`\n   switches to replace).\n3. Kit default: `assets\u002Farchetype-catalog-mapping.yaml` shipped with\n   the team-creator skill.\n\nValidate eagerly: every archetype entry must declare a `role` that\nstarts with `ROLE-` and a non-empty `seniority`. Replace-mode\noverrides must list all 8 archetypes (missing archetypes are a hard\nerror).\n\nRecord:\n- `inputs_snapshot.archetype_catalog_mapping_file`: `\"kit-default\"`,\n  `\"project\"`, or `\"cli\"` — which layer was applied.\n- `inputs_snapshot.archetype_catalog_overrides`: per-archetype\n  per-field deltas relative to the kit default. Empty list when the\n  kit default was used verbatim.\n\n**Layer B — role-archetypes override (eager validation — before archetype mapping):**\n\nIf `context\u002Fteam\u002Frole-archetypes.yaml` exists, load it now and\nvalidate it immediately against the rules in\n`references\u002Frole-archetypes-override.md` (§Validation invariants).\nValidation failures are hard errors — abort before any mapping begins.\nThis includes: PM pin must remain heavy, all rationales non-empty,\nall 8 archetypes present in `replace` mode. (The legacy\n`clone_cap_override` field is no longer accepted — capacity is the\ncount of RoleSlots opened in step 6 below.)\n\nRecord `inputs_snapshot.archetype_override_file: \"present\"` (or\n`\"absent\"`) and list each overridden role with its new pin and\nrationale summary in `inputs_snapshot.archetype_overrides`.\n\n**Archetype mapping:**\n\nFor each of the 8 role archetypes:\n\n1. Select the archetype's pinned tier — from the (possibly-overridden)\n   archetype table. Kit defaults (`references\u002Frole-archetypes.md`)\n   apply for any archetype not listed in a delta override, or when no\n   override file is present.\n2. From the models classified into that tier, pick the highest TierScore.\n3. If preferred_providers are supplied and two models score within 0.05\n   of each other, prefer the model from the preferred provider.\n4. If the pinned tier has no candidates: apply the override-when rule\n   from the active archetype table; if no override applies, fail\n   with a clear message.\n5. Resolve the archetype's `(role, seniority)` pair via the mapping\n   from Layer A. Verify `role` is present in `context\u002Froles\u002F`; fail\n   with a clear message if the catalog Role file is missing.\n\n### Step 5 — Deactivate prior team (if re-running)\n\nIf `context\u002Fteam\u002Froster.md` already exists:\n\n1. Parse the prior team's RoleSlot and Binding IDs from roster.md.\n2. For each prior `role-slot-fill` Binding:\n   `binding-management.end_binding(id, reason=\n     \"superseded by pk-team-create run \u003CISO-timestamp>\")`\n3. For each prior RoleSlot under the prior chartering Scope:\n   `team-manager.close_role_slot(id, reason=\n     \"superseded by pk-team-create run \u003CISO-timestamp>\")`.\n   The slot state machine forbids reopening — re-charters always\n   emit fresh `SLOT-*` IDs under the new chartering Scope. There is\n   no Actor-template \u002F Actor-clone distinction in v2; capacity is\n   expressed by RoleSlot count, not by clone Actors.\n\nIf the prior team predates v2 (its roster references v1\n`role-assignment` Bindings and archetype Roles), do NOT touch those\nentities. The Phase A migration apply script\n(`team-manager\u002Fscripts\u002Fapply_migration_2139.py`) is responsible for\nback-filling the v1 surface; pk-team-create only operates on the v2\nRoleSlot surface.\n\n### Step 6 — Write entities\n\nSkip if `--dry-run`.\n\nFor each of the 8 archetypes, look up `(role, seniority)` in the\nmapping resolved in step 4. Open one RoleSlot per parallelism unit\n(rank 1..N, where N is `--parallelism-cap`; `project-manager` is\nalways exactly 1 slot at rank=1):\n\n```\nteam-manager.create_role_slot(\n  scope=\u003C--chartering-scope>,\n  role=\u003CROLE-id from mapping>,\n  seniority=\u003Cseniority from mapping>,\n  rank=\u003Cint 1..N>,\n  rationale=\"archetype=\u003Carchetype-name> tier=\u003Ctier> \"\n            \"score=\u003Ctier-score> model=\u003Cmodel-id>\",\n  default_model_profile=\u003CART-*-ModelProfile-*>     # optional\n)\n```\n\nThen, for each TeamMember selected to fill a slot:\n\n```\nteam-manager.fill_role_slot(\n  id=\u003CSLOT-id>,\n  team_member_slug=\u003Cteam-member-slug>,\n  rationale=\"\u003Cwhy this person for this slot>\",\n  valid_from=\u003Ctoday>,\n  valid_until=\u003Cchartering Scope's ends_at, if any>\n)\n```\n\n`fill_role_slot` writes the `role-slot-fill` Binding inline. No\narchetype Role, Actor, or `role-assignment` Binding is written by\npk-team-create.\n\nEphemeral archetypes that have no persistent TeamMember to assign\n(e.g. `assistant`, when handled by ad-hoc dispatch) leave the slot\nin `state=open` with a `default_model_profile` set; the resolver\npre-step in `team-manager.get_interlocutor_runtime_binding` will\nreturn the open slot's profile for ephemeral lookups.\n\n### Step 7 — Write roster.md\n\nWrite `context\u002Fteam\u002Froster.md` with:\n- Narrative preamble (team charter, subscription, governance floor)\n- Routing table: role → actor → model-id → tier → tier-score\n- Parallelism cap and PM clone-cap note\n- Weights used (from inputs or defaults)\n- Landscape artifact ID and date\n- DecisionRecord ID (written in step 8)\n\n### Step 7.5 — Compute budget projection\n\nAfter step 6 writes all RoleSlots and before writing the chartering\nDecisionRecord, iterate every created slot and compute a cost projection.\n\nFor each RoleSlot:\n\n1. Call `model-recommender.get_pricing(\u003Cmodel_profile>)` to fetch the\n   live unit cost.  Snapshot the returned `price_per_token_usd` into\n   `unit_cost_usd` in the projection row.\n2. Determine the effective time window for the slot:\n   - If the slot is filled by a `type=consultant` TeamMember: intersect\n     the consultant's `engagement_window` with the chartering Scope's\n     `{starts_at, ends_at}`. Use the intersected window.\n   - Otherwise: use the chartering Scope's full window.\n3. Apply heuristic formula (when `--projection-method heuristic`):\n   ```\n   weeks = ceil(effective_window_days \u002F 7)\n   projected_total_usd = unit_cost_usd\n                         × (avg_input_tokens + avg_output_tokens)\n                         × expected_invocations_per_week\n                         × weeks\n   ```\n   Defaults when not pinned by the landscape artifact:\n   - `expected_invocations_per_week`: 50\n   - `avg_tokens_per_invocation.input`: 8000\n   - `avg_tokens_per_invocation.output`: 2000\n\n4. Sum slot `projected_total_usd` values into `projected_total`.\n\n5. Build the `budget_projection` block (see shape in step 8 below).\n   Pass `drift_threshold_pct` from `--budget-drift-threshold` (default 20).\n\nSkip budget projection if the chartering Scope has no `starts_at`\u002F`ends_at`\ndates; log a warning and set `budget_projection: null` in the snapshot.\n\n### Step 8 — Write chartering DecisionRecord\n\n```\ndecision-record-write(\n  title=\"Team composition — \u003Csubscription> — \u003Cdate>\",\n  state=\"accepted\",\n  supersedes=[\u003Cprior-team-decisionrecord-id>],  # if re-running\n  inputs_snapshot={\n    subscription: ...,\n    governance_floor: ...,\n    parallelism_cap: ...,\n    weights: {C, K, L, G},\n    weights_source: \"cli\" | \"dec-team-weights\" | \"skill-default\",\n    tier_thresholds: {heavy_min, medium_min},\n    tier_thresholds_source: \"cli\" | \"dec-team-weights\" | \"skill-default\",\n    landscape_artifact: \u003CART-id>,\n    landscape_artifact_date: \u003Cdate>,\n    landscape_artifact_source: \"explicit\" | \"project-tag\" | \"kit-default\",\n    tier_scores: {\u003Cmodel-id>: \u003Cscore>, ...},\n    weight_overrides_applied: \u003Cbool>,\n    # v2 catalog-mapping audit (additionalProperties=true on inputs_snapshot)\n    archetype_catalog_mapping_file:\n      \"kit-default\" | \"project\" | \"cli\",\n    archetype_catalog_overrides: [\n      {archetype, field, kit_default, project_value}, ...\n    ],\n    archetype_override_file: \"present\" | \"absent\",\n    archetype_override_semantics: \"delta\" | \"replace\" | null,\n    archetype_overrides: [...],   # list of {role, kit_default_pin, override_pin}\n    chartering_scope: \u003CSCOPE-id>,\n    role_slots: [\n      {archetype, slot_id, role, seniority, rank}, ...\n    ],\n    # Gap 5 — budget projection (additionalProperties=true; additive)\n    budget_projection: {\n      currency: USD,\n      window: {starts_at: \u003CISO>, ends_at: \u003CISO>},  # = chartering Scope window\n      projected_total: \u003Cfloat>,\n      projection_method: heuristic | model-recommender-quote | manual,\n      slot_projections: [\n        {\n          slot: SLOT-\u003Cid>,\n          role: ROLE-\u003Cid>,\n          seniority: \u003Cenum>,\n          model_profile: ART-*-ModelProfile-*,\n          expected_invocations_per_week: \u003Cint>,\n          avg_tokens_per_invocation: {input: \u003Cint>, output: \u003Cint>},\n          unit_cost_usd: \u003Cfloat>,          # snapshotted from get_pricing at charter time\n          projected_total_usd: \u003Cfloat>,\n          effective_window: {starts_at: \u003CISO>, ends_at: \u003CISO>},\n        }, ...\n      ],\n      drift_threshold_pct: 20,             # from --budget-drift-threshold, default 20\n      notes: \u003Cfree-text or null>,\n    }\n  }\n)\n```\n\nThe `inputs_snapshot.weights` block is the canonical weight store\nthat `pk-team-rebalance` will read on future runs. The\n`chartering_scope` and `role_slots` blocks are the v2 provenance\nback-pointer from the DEC to the RoleSlots opened in step 6.\n\nThe `budget_projection` block is Gap 5 (SUB-4 \u002F SwiftReef). It is\nadditive (`additionalProperties: true` on the decision_record schema);\nold DecisionRecords without this block validate cleanly. The block is\nthe source of truth that `pk-team-review` Step 5c reads for drift\ndetection.\n\n## Dry-run output format\n\n```\n=== pk-team-create DRY RUN ===\nSubscription: \u003Cvalue>   Governance floor: \u003Cvalue>\nChartering Scope: \u003CSCOPE-id>\nLandscape: \u003Cartifact-id> (\u003Cdate>)\nWeights: C=0.60 K=0.20 L=0.10 G=0.10\nMapping source: kit-default | project | cli   (overrides: \u003Ccount>)\n\nCandidate models scored:\n  \u003Cmodel-id>  TierScore=0.92  → heavy\n  \u003Cmodel-id>  TierScore=0.61  → medium\n  ...\n\nRoleSlot plan (per archetype):\n  project-manager  → ROLE-product-manager \u002F senior  (heavy, score=0.92)\n    1 slot at rank=1, fill: TEAMMEMBER-\u003Cslug>, model=\u003Cmodel-id>\n  senior-architect → ROLE-solutions-architect \u002F senior  (heavy, score=0.87)\n    N slots at rank=1..N, fill: TEAMMEMBER-\u003Cslug>, model=\u003Cmodel-id>\n  ...\n\nEntities to write (skipped in dry-run):\n  \u003Ctotal RoleSlot count> RoleSlot entities\n  \u003Ctotal fill count>     role-slot-fill Binding entities\n  context\u002Fteam\u002Froster.md\n  DecisionRecord superseding \u003CDEC-id>\n===========================\n```\n\n## State side-effects (non-dry-run)\n\nCreates: one RoleSlot per parallelism unit (under the chartering\nScope), one `role-slot-fill` Binding per filled slot, roster.md,\n1 DecisionRecord. Closes prior RoleSlots and ends prior\n`role-slot-fill` Bindings on re-charter. Does NOT write archetype\nRole entities, Actor entities, or `role-assignment` Bindings —\nthose v1 surfaces are read-only during the deprecation window.\n\n\n---\n\nThis command is a processkit skill shim. Load and follow the matching skill for `pk-team-create` from `context\u002Fskills\u002F` instead of executing underlying helper scripts directly. Do not run `context\u002Fskills\u002F**\u002Fscripts\u002F*.py`, `doctor.py`, or `uv run ...\u002Fscripts\u002F...` unless the skill instructions explicitly require that implementation detail for the current step.\n","# Command: pk-team-create\n\nFull team derivation from scratch (v2 \u002F catalog-driven). Selects\ncatalog Roles via the archetype-catalog mapping, opens RoleSlots\nunder the chartering Scope, fills them with TeamMembers, rewrites\n`context\u002Fteam\u002Froster.md`, and emits a chartering DecisionRecord that\nsupersedes the prior team DecisionRecord. **No archetype Role\nentities are written.**\n\n## Syntax\n\n```\npk-team-create\n  --chartering-scope \u003CSCOPE-id>      # required — Scope that owns the team's RoleSlots\n  --subscription \u003Cprovider>:\u003Ctier>   # e.g. anthropic:max-5x\n  --providers \u003Clist|\"any\">           # comma-separated; \"any\" = all accessible\n  --parallelism-cap \u003Cint>            # max RoleSlots opened per archetype, default 5\n  --governance-floor \u003C0-5>           # G-score floor, default 3\n  [--weight-overrides \u003Cjson>]        # {\"C\":0.60,\"K\":0.20,\"L\":0.10,\"G\":0.10}\n                                     # CLI > DEC-*-TeamWeights > skill defaults\n  [--threshold-overrides \u003Cjson>]     # {\"heavy_min\":0.70,\"medium_min\":0.40}\n                                     # CLI > DEC-*-TeamWeights > skill defaults\n  [--landscape-artifact \u003CART-id>]    # explicit landscape override; skips discovery\n  [--landscape \u003Cartifact-id>]        # alias for --landscape-artifact (kept for compat)\n  [--archetype-catalog-mapping \u003Cpath>]\n                                     # CLI override of the archetype→catalog\n                                     # mapping (replace mode). Project override\n                                     # at context\u002Fteam\u002Farchetype-catalog-mapping.yaml\n                                     # is auto-detected; this flag wins over both.\n  [--budget-drift-threshold \u003Cfloat>] # drift alert threshold %, default 20\n  [--projection-method \u003Cmethod>]     # heuristic | model-recommender-quote | manual\n                                     # default: heuristic\n  [--dry-run]                        # print plan; write nothing\n```\n\n## Process (sequential)\n\n### Step 1 — Resolve landscape snapshot\n\nThree-level precedence (see `references\u002Flandscape-resolution.md`):\n\n1. If `--landscape-artifact \u003CART-id>` (or alias `--landscape`) is\n   supplied, load that artifact directly. Record\n   `inputs_snapshot.landscape_artifact_source: \"explicit\"`.\n2. Else query artifact index for the most-recently-created artifact\n   tagged both `landscape-summary` **and** `landscape-summary-project`\n   whose `spec.project_id` matches the current project context.\n   Record `inputs_snapshot.landscape_artifact_source: \"project-tag\"`.\n3. Else fall back to the most-recently-created artifact tagged\n   `landscape-summary` (kit default). Record\n   `inputs_snapshot.landscape_artifact_source: \"kit-default\"`.\n\nIf no artifact is found at any level, **abort** with:\n> \"No landscape-summary artifact found. Run the Haiku landscape-ingest\n> task to produce an ART-*-LandscapeSnapshot artifact before running\n> pk-team-create.\"\n\nIf the resolved artifact is older than 90 days, emit a warning but\ncontinue. Record the artifact ID and its `created` date in the\nDecisionRecord `inputs_snapshot`.\n\n### Step 2 — Query accessible models\n\n```\nmodel-recommender.check_availability()\nmodel-recommender.query_models(\n  providers=\u003C--providers>,\n  G_floor=\u003C--governance-floor>,\n  apply_user_filter=true\n)\nmodel-recommender.get_pricing(sort_by=\"value_score\")\n```\n\nFilter out any model with availability != \"operational\" unless the\nowner explicitly passes `--allow-degraded`.\n\n### Step 3 — Resolve weights + thresholds, then apply tiering formula\n\n**Weight resolution (before scoring):**\n\n```\nCLI --weight-overrides \u003Cjson>\n  > DEC-*-TeamWeights.weight_overrides (tag: team-weights-override,\n      most recent accepted; see references\u002Fteam-weights-decision-schema.md)\n    > skill defaults: {C:0.60, K:0.20, L:0.10, G:0.10}\n```\n\n**Threshold resolution (before classification):**\n\n```\nCLI --threshold-overrides \u003Cjson>\n  > DEC-*-TeamWeights.tier_thresholds (same DEC; validated on load)\n    > skill defaults: {heavy_min:0.70, medium_min:0.40}\n```\n\nValidate any threshold override against the rules in\n`references\u002Fteam-weights-decision-schema.md` (§Threshold validation\nrules). Violations are hard errors — abort before scoring.\n\nRecord `inputs_snapshot.weights_source` and\n`inputs_snapshot.tier_thresholds_source` with the resolution level\nused (`\"cli\"`, `\"dec-team-weights\"`, or `\"skill-default\"`), and\nstore the actual values in `inputs_snapshot.weights` and\n`inputs_snapshot.tier_thresholds`.\n\n**Scoring:**\n\nFor each accessible candidate model `m`:\n\n1. Compute raw dimension values:\n   - `C_raw(m)` = (swe_bench_verified + swe_bench_pro +\n                   community_rating) \u002F 3\n   - `K_raw(m)` = 1 \u002F (output_price_per_1M × quota_burn_vs_flagship)\n   - `L_raw(m)` = output_tokens_per_sec\n   - `G_raw(m)` = G-score from model-recommender (0–5 scale,\n                  normalise to 0–1 by dividing by 5)\n\n2. Min-max normalise each dimension across the full candidate set:\n   `norm(x) = clamp((x - min) \u002F (max - min), 0.01, 1.00)`\n\n3. Compute TierScore using resolved weights `{C, K, L, G}`:\n   `TierScore(m) = C×norm(C) + K×norm(K) + L×norm(L) + G×norm(G)`\n\n4. Classify using resolved thresholds:\n   - TierScore ≥ `heavy_min` → **heavy**\n   - TierScore ≥ `medium_min` and \u003C `heavy_min` → **medium**\n   - TierScore \u003C `medium_min` → **light**\n\n5. Tier-collapse rule: if fewer than 3 distinct tiers result, promote\n   the two highest-scoring light models to medium.\n\nSee `references\u002Ftiering-formula.md` for the full formula and worked\nexample.\n\n### Step 4 — Load archetype-catalog mapping + role-archetypes override, then map archetypes\n\n**Layer A — archetype-catalog mapping (eager validation — before archetype mapping):**\n\nResolve the archetype → catalog `(ROLE-id, seniority)` mapping with\nthree-level precedence:\n\n1. `--archetype-catalog-mapping \u003Cpath>` flag (CLI; replace semantics).\n2. `context\u002Fteam\u002Farchetype-catalog-mapping.yaml` (project; delta\n   semantics by default — top-level `override_semantics: replace`\n   switches to replace).\n3. Kit default: `assets\u002Farchetype-catalog-mapping.yaml` shipped with\n   the team-creator skill.\n\nValidate eagerly: every archetype entry must declare a `role` that\nstarts with `ROLE-` and a non-empty `seniority`. Replace-mode\noverrides must list all 8 archetypes (missing archetypes are a hard\nerror).\n\nRecord:\n- `inputs_snapshot.archetype_catalog_mapping_file`: `\"kit-default\"`,\n  `\"project\"`, or `\"cli\"` — which layer was applied.\n- `inputs_snapshot.archetype_catalog_overrides`: per-archetype\n  per-field deltas relative to the kit default. Empty list when the\n  kit default was used verbatim.\n\n**Layer B — role-archetypes override (eager validation — before archetype mapping):**\n\nIf `context\u002Fteam\u002Frole-archetypes.yaml` exists, load it now and\nvalidate it immediately against the rules in\n`references\u002Frole-archetypes-override.md` (§Validation invariants).\nValidation failures are hard errors — abort before any mapping begins.\nThis includes: PM pin must remain heavy, all rationales non-empty,\nall 8 archetypes present in `replace` mode. (The legacy\n`clone_cap_override` field is no longer accepted — capacity is the\ncount of RoleSlots opened in step 6 below.)\n\nRecord `inputs_snapshot.archetype_override_file: \"present\"` (or\n`\"absent\"`) and list each overridden role with its new pin and\nrationale summary in `inputs_snapshot.archetype_overrides`.\n\n**Archetype mapping:**\n\nFor each of the 8 role archetypes:\n\n1. Select the archetype's pinned tier — from the (possibly-overridden)\n   archetype table. Kit defaults (`references\u002Frole-archetypes.md`)\n   apply for any archetype not listed in a delta override, or when no\n   override file is present.\n2. From the models classified into that tier, pick the highest TierScore.\n3. If preferred_providers are supplied and two models score within 0.05\n   of each other, prefer the model from the preferred provider.\n4. If the pinned tier has no candidates: apply the override-when rule\n   from the active archetype table; if no override applies, fail\n   with a clear message.\n5. Resolve the archetype's `(role, seniority)` pair via the mapping\n   from Layer A. Verify `role` is present in `context\u002Froles\u002F`; fail\n   with a clear message if the catalog Role file is missing.\n\n### Step 5 — Deactivate prior team (if re-running)\n\nIf `context\u002Fteam\u002Froster.md` already exists:\n\n1. Parse the prior team's RoleSlot and Binding IDs from roster.md.\n2. For each prior `role-slot-fill` Binding:\n   `binding-management.end_binding(id, reason=\n     \"superseded by pk-team-create run \u003CISO-timestamp>\")`\n3. For each prior RoleSlot under the prior chartering Scope:\n   `team-manager.close_role_slot(id, reason=\n     \"superseded by pk-team-create run \u003CISO-timestamp>\")`.\n   The slot state machine forbids reopening — re-charters always\n   emit fresh `SLOT-*` IDs under the new chartering Scope. There is\n   no Actor-template \u002F Actor-clone distinction in v2; capacity is\n   expressed by RoleSlot count, not by clone Actors.\n\nIf the prior team predates v2 (its roster references v1\n`role-assignment` Bindings and archetype Roles), do NOT touch those\nentities. The Phase A migration apply script\n(`team-manager\u002Fscripts\u002Fapply_migration_2139.py`) is responsible for\nback-filling the v1 surface; pk-team-create only operates on the v2\nRoleSlot surface.\n\n### Step 6 — Write entities\n\nSkip if `--dry-run`.\n\nFor each of the 8 archetypes, look up `(role, seniority)` in the\nmapping resolved in step 4. Open one RoleSlot per parallelism unit\n(rank 1..N, where N is `--parallelism-cap`; `project-manager` is\nalways exactly 1 slot at rank=1):\n\n```\nteam-manager.create_role_slot(\n  scope=\u003C--chartering-scope>,\n  role=\u003CROLE-id from mapping>,\n  seniority=\u003Cseniority from mapping>,\n  rank=\u003Cint 1..N>,\n  rationale=\"archetype=\u003Carchetype-name> tier=\u003Ctier> \"\n            \"score=\u003Ctier-score> model=\u003Cmodel-id>\",\n  default_model_profile=\u003CART-*-ModelProfile-*>     # optional\n)\n```\n\nThen, for each TeamMember selected to fill a slot:\n\n```\nteam-manager.fill_role_slot(\n  id=\u003CSLOT-id>,\n  team_member_slug=\u003Cteam-member-slug>,\n  rationale=\"\u003Cwhy this person for this slot>\",\n  valid_from=\u003Ctoday>,\n  valid_until=\u003Cchartering Scope's ends_at, if any>\n)\n```\n\n`fill_role_slot` writes the `role-slot-fill` Binding inline. No\narchetype Role, Actor, or `role-assignment` Binding is written by\npk-team-create.\n\nEphemeral archetypes that have no persistent TeamMember to assign\n(e.g. `assistant`, when handled by ad-hoc dispatch) leave the slot\nin `state=open` with a `default_model_profile` set; the resolver\npre-step in `team-manager.get_interlocutor_runtime_binding` will\nreturn the open slot's profile for ephemeral lookups.\n\n### Step 7 — Write roster.md\n\nWrite `context\u002Fteam\u002Froster.md` with:\n- Narrative preamble (team charter, subscription, governance floor)\n- Routing table: role → actor → model-id → tier → tier-score\n- Parallelism cap and PM clone-cap note\n- Weights used (from inputs or defaults)\n- Landscape artifact ID and date\n- DecisionRecord ID (written in step 8)\n\n### Step 7.5 — Compute budget projection\n\nAfter step 6 writes all RoleSlots and before writing the chartering\nDecisionRecord, iterate every created slot and compute a cost projection.\n\nFor each RoleSlot:\n\n1. Call `model-recommender.get_pricing(\u003Cmodel_profile>)` to fetch the\n   live unit cost.  Snapshot the returned `price_per_token_usd` into\n   `unit_cost_usd` in the projection row.\n2. Determine the effective time window for the slot:\n   - If the slot is filled by a `type=consultant` TeamMember: intersect\n     the consultant's `engagement_window` with the chartering Scope's\n     `{starts_at, ends_at}`. Use the intersected window.\n   - Otherwise: use the chartering Scope's full window.\n3. Apply heuristic formula (when `--projection-method heuristic`):\n   ```\n   weeks = ceil(effective_window_days \u002F 7)\n   projected_total_usd = unit_cost_usd\n                         × (avg_input_tokens + avg_output_tokens)\n                         × expected_invocations_per_week\n                         × weeks\n   ```\n   Defaults when not pinned by the landscape artifact:\n   - `expected_invocations_per_week`: 50\n   - `avg_tokens_per_invocation.input`: 8000\n   - `avg_tokens_per_invocation.output`: 2000\n\n4. Sum slot `projected_total_usd` values into `projected_total`.\n\n5. Build the `budget_projection` block (see shape in step 8 below).\n   Pass `drift_threshold_pct` from `--budget-drift-threshold` (default 20).\n\nSkip budget projection if the chartering Scope has no `starts_at`\u002F`ends_at`\ndates; log a warning and set `budget_projection: null` in the snapshot.\n\n### Step 8 — Write chartering DecisionRecord\n\n```\ndecision-record-write(\n  title=\"Team composition — \u003Csubscription> — \u003Cdate>\",\n  state=\"accepted\",\n  supersedes=[\u003Cprior-team-decisionrecord-id>],  # if re-running\n  inputs_snapshot={\n    subscription: ...,\n    governance_floor: ...,\n    parallelism_cap: ...,\n    weights: {C, K, L, G},\n    weights_source: \"cli\" | \"dec-team-weights\" | \"skill-default\",\n    tier_thresholds: {heavy_min, medium_min},\n    tier_thresholds_source: \"cli\" | \"dec-team-weights\" | \"skill-default\",\n    landscape_artifact: \u003CART-id>,\n    landscape_artifact_date: \u003Cdate>,\n    landscape_artifact_source: \"explicit\" | \"project-tag\" | \"kit-default\",\n    tier_scores: {\u003Cmodel-id>: \u003Cscore>, ...},\n    weight_overrides_applied: \u003Cbool>,\n    # v2 catalog-mapping audit (additionalProperties=true on inputs_snapshot)\n    archetype_catalog_mapping_file:\n      \"kit-default\" | \"project\" | \"cli\",\n    archetype_catalog_overrides: [\n      {archetype, field, kit_default, project_value}, ...\n    ],\n    archetype_override_file: \"present\" | \"absent\",\n    archetype_override_semantics: \"delta\" | \"replace\" | null,\n    archetype_overrides: [...],   # list of {role, kit_default_pin, override_pin}\n    chartering_scope: \u003CSCOPE-id>,\n    role_slots: [\n      {archetype, slot_id, role, seniority, rank}, ...\n    ],\n    # Gap 5 — budget projection (additionalProperties=true; additive)\n    budget_projection: {\n      currency: USD,\n      window: {starts_at: \u003CISO>, ends_at: \u003CISO>},  # = chartering Scope window\n      projected_total: \u003Cfloat>,\n      projection_method: heuristic | model-recommender-quote | manual,\n      slot_projections: [\n        {\n          slot: SLOT-\u003Cid>,\n          role: ROLE-\u003Cid>,\n          seniority: \u003Cenum>,\n          model_profile: ART-*-ModelProfile-*,\n          expected_invocations_per_week: \u003Cint>,\n          avg_tokens_per_invocation: {input: \u003Cint>, output: \u003Cint>},\n          unit_cost_usd: \u003Cfloat>,          # snapshotted from get_pricing at charter time\n          projected_total_usd: \u003Cfloat>,\n          effective_window: {starts_at: \u003CISO>, ends_at: \u003CISO>},\n        }, ...\n      ],\n      drift_threshold_pct: 20,             # from --budget-drift-threshold, default 20\n      notes: \u003Cfree-text or null>,\n    }\n  }\n)\n```\n\nThe `inputs_snapshot.weights` block is the canonical weight store\nthat `pk-team-rebalance` will read on future runs. The\n`chartering_scope` and `role_slots` blocks are the v2 provenance\nback-pointer from the DEC to the RoleSlots opened in step 6.\n\nThe `budget_projection` block is Gap 5 (SUB-4 \u002F SwiftReef). It is\nadditive (`additionalProperties: true` on the decision_record schema);\nold DecisionRecords without this block validate cleanly. The block is\nthe source of truth that `pk-team-review` Step 5c reads for drift\ndetection.\n\n## Dry-run output format\n\n```\n=== pk-team-create DRY RUN ===\nSubscription: \u003Cvalue>   Governance floor: \u003Cvalue>\nChartering Scope: \u003CSCOPE-id>\nLandscape: \u003Cartifact-id> (\u003Cdate>)\nWeights: C=0.60 K=0.20 L=0.10 G=0.10\nMapping source: kit-default | project | cli   (overrides: \u003Ccount>)\n\nCandidate models scored:\n  \u003Cmodel-id>  TierScore=0.92  → heavy\n  \u003Cmodel-id>  TierScore=0.61  → medium\n  ...\n\nRoleSlot plan (per archetype):\n  project-manager  → ROLE-product-manager \u002F senior  (heavy, score=0.92)\n    1 slot at rank=1, fill: TEAMMEMBER-\u003Cslug>, model=\u003Cmodel-id>\n  senior-architect → ROLE-solutions-architect \u002F senior  (heavy, score=0.87)\n    N slots at rank=1..N, fill: TEAMMEMBER-\u003Cslug>, model=\u003Cmodel-id>\n  ...\n\nEntities to write (skipped in dry-run):\n  \u003Ctotal RoleSlot count> RoleSlot entities\n  \u003Ctotal fill count>     role-slot-fill Binding entities\n  context\u002Fteam\u002Froster.md\n  DecisionRecord superseding \u003CDEC-id>\n===========================\n```\n\n## State side-effects (non-dry-run)\n\nCreates: one RoleSlot per parallelism unit (under the chartering\nScope), one `role-slot-fill` Binding per filled slot, roster.md,\n1 DecisionRecord. Closes prior RoleSlots and ends prior\n`role-slot-fill` Bindings on re-charter. Does NOT write archetype\nRole entities, Actor entities, or `role-assignment` Bindings —\nthose v1 surfaces are read-only during the deprecation window.\n\n\n---\n\nThis command is a processkit skill shim. Load and follow the matching skill for `pk-team-create` from `context\u002Fskills\u002F` instead of executing underlying helper scripts directly. Do not run `context\u002Fskills\u002F**\u002Fscripts\u002F*.py`, `doctor.py`, or `uv run ...\u002Fscripts\u002F...` unless the skill instructions explicitly require that implementation detail for the current step.\n",{"name":5,"description":7},"main","4876a92620ee792aac2ac9a1a806322d8a572367","archive\u002Fprocesskit-install-20261101\u002Fharness\u002Fclaude-skills\u002Fpk-team-create","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":36,"prompt":46,"zipAvailable":47},[37,40,43],{"tool":38,"command":39},"npx","npx skills add projectious-work\u002Fprocesskit --skill pk-team-create",{"tool":41,"command":42},"bunx","bunx skills add projectious-work\u002Fprocesskit --skill pk-team-create",{"tool":44,"command":45},"pnpm","pnpm dlx skills add projectious-work\u002Fprocesskit --skill pk-team-create","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fprojectious-work\u002Fprocesskit\u002Fblob\u002Fmain\u002Farchive\u002Fprocesskit-install-20261101\u002Fharness\u002Fclaude-skills\u002Fpk-team-create\u002FSKILL.md",true,[],[50,57,64,70,97,117],{"skillNo":51,"name":52,"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":53,"categories":54,"keywords":55,"publishedAt":56},"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":58,"name":59,"nameZh":6,"description":60,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":61,"categories":62,"keywords":63,"publishedAt":56},"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":65,"name":66,"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":67,"categories":68,"keywords":69,"publishedAt":56},"01M76650NN4T8P241021NBR64S","standard-agent-building","https:\u002F\u002Fgithub.com\u002Fprojectious-work\u002Fprocesskit\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Fstandard-agent-building\u002FSKILL.md",[],[],{"skillNo":71,"name":72,"nameZh":6,"description":73,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":74,"categories":75,"keywords":81,"publishedAt":23},"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",[76],{"categoryNo":77,"categoryName":78,"taxonomyType":79,"taxonomyTypeDesc":80},"01JBSEEDCAT000000000000023","Skill与智能体开发","SCENARIO","使用场景",[82,83,84,85,19,86,87,88,89,90,91,92,72,93,94,95,96],"management","skill management","ontology","v1","as","work","transition","get","manage","update","alpha","list","package","skill package","manifests",{"skillNo":98,"name":99,"nameZh":6,"description":100,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":101,"categories":102,"keywords":104,"publishedAt":23},"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",[103],{"categoryNo":77,"categoryName":78,"taxonomyType":79,"taxonomyTypeDesc":80},[105,106,82,107,108,109,110,111,112,113,114,115,87,89,90,92,116],"schema","schema management","contract","rebuild","contracts","workflows","wed","generated","MCP","validation","deterministic","mode",{"skillNo":118,"name":119,"nameZh":6,"description":120,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":121,"categories":122,"keywords":123,"publishedAt":23},"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",[],[124,82,125,126,19,127,128,129,87,89,130,131,91,132,92,133,134,135,136,137],"proposition management","proposition","interface","claims","record","world","creation","event","WSJF","index","updates","query","facts","log"]