[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M71RFW0T2SEPXGHEQR954XT9":3,"skill-same:01M71RFW0T2SEPXGHEQR954XT9":95},{"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":23,"publishedAt":44,"linkOnly":45,"markdownContent":46,"markdownBody":47,"frontMatter":48,"refName":49,"commitSha":50,"skillDirectory":51,"forksCount":52,"repositoryHtmlUrl":53,"repositoryHomepageUrl":6,"ownerHtmlUrl":54,"ownerBlogUrl":6,"ownerDescription":55,"updatedAt":56,"install":57,"resources":70},"01M71RFW0T2SEPXGHEQR954XT9","autospec-persona",null,"Use when the operator wants to calibrate autospec to their judgment — runs a repo-grounded calibration interview (≤50 questions in themed AskUserQuestion batches), interleaves calibration questions whose answers are inferable from memory\u002Fcharter to measure agreement, and writes ~\u002F.autospec\u002Foperator-persona.answers.json. Resumable — re-running continues at the next pending batch and never re-asks done batches.","berlinguyinca\u002Fautospec","berlinguyinca","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F16364?v=4","01M71R7C0SWBG05FNHQE4XQ7Y3","Gert Wohlgemuth","01M71R7C1VY5K3Q715RNDDBVGQ",3,"MIT","https:\u002F\u002Fgithub.com\u002Fberlinguyinca\u002Fautospec\u002Fblob\u002Fmain\u002Fskills\u002Fautospec-persona\u002FSKILL.md",[18],{"categoryNo":19,"categoryName":20,"taxonomyType":21,"taxonomyTypeDesc":22},"01JBSEEDCAT000000000000038","人力资源","PROFESSION","适用职业",[24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43],"persona","autospec","autospec persona","charter","next","batch","agreement","batches","done","answers","pending","when","use","re","calibrate","repo","memory","never","interview","running","2026-11-10 13:02:02",false,"---\nname: autospec-persona\ndescription: Use when the operator wants to calibrate autospec to their judgment — runs a repo-grounded calibration interview (≤50 questions in themed AskUserQuestion batches), interleaves calibration questions whose answers are inferable from memory\u002Fcharter to measure agreement, and writes ~\u002F.autospec\u002Foperator-persona.answers.json. Resumable — re-running continues at the next pending batch and never re-asks done batches.\n---\n\n# autospec-persona (harness-neutral)\n\nStandalone calibration-interview skill. A Tier-A agent reads the current repo\n(`docs\u002Fspecs\u002F`, `docs\u002Fmemory\u002F`, `AGENTS.md`, the autonomy charter, recent\ncommits, open issues, and the main code areas), surfaces the real tensions and\nrecurring decisions, and turns them into **≤50 repo-grounded questions** (a hard\ncap) asked in themed `AskUserQuestion` batches (multiple-choice preferred). The\noperator's answers persist to `~\u002F.autospec\u002Foperator-persona.answers.json` and the\nrun is fully **resumable**: re-invoking continues at the next pending batch and\nnever re-asks a `done` batch.\n\nManage your own context — never exceed 60%. Delegate to subagents whenever your harness supports it.\n\n\u003C!-- autospec-block:startup-self-update SKILL_NAME=autospec-persona -->\n\n## Self-update mode\n\nIf the feature-request argument matches the regex `^\\s*update\\s*$` (case-insensitive, whitespace-padded), this skill enters self-update mode and does not run the normal interview:\n\n1. **Detect harness** by checking which install path exists for this skill:\n   - Claude Code: `~\u002F.claude\u002Fskills\u002Fautospec-persona\u002FSKILL.md`\n   - OpenCode:    `~\u002F.config\u002Fopencode\u002Fagent\u002Fautospec-persona.md`\n   - Codex CLI:   `~\u002F.codex\u002Fprompts\u002Fautospec-persona.md`\n2. **Re-install the full autospec suite from `main`** by piping the canonical installer:\n   ```bash\n   curl -fsSL https:\u002F\u002Fraw.githubusercontent.com\u002Fberlinguyinca\u002Fautospec\u002Fmain\u002Fbootstrap.sh | bash -s -- --skill all --harness all --update\n   ```\n   Run this one-liner once; it refreshes all autospec skills across all harnesses.\n3. **Show the diff** between the prior installed file(s) and the freshly fetched copy (e.g. `diff \u003C(cat \u003Cprior>) \u003C(curl -fsSL ...SKILL.md)` or the equivalent recorded by the installer).\n4. **Stop.** Do not read the repo, generate questions, or run the interview. Print the upgrade summary and return to the user.\n\nIf no install path is detected, print `Self-update: no installed copy of autospec-persona found; run install.sh first.` and exit.\n\n## Invocation\n\n```\n\u002Fautospec-persona [--reset] [--max-questions N] [--dry-run]\n```\n\n- `--reset` — discard any existing `~\u002F.autospec\u002Foperator-persona.answers.json` and\n  start a fresh interview from `next_batch=0`. Without it, the skill **resumes**.\n- `--max-questions N` — lower the question budget below the hard cap of 50\n  (never above; a value > 50 is clamped to 50).\n- `--dry-run` — generate and print the themed batches and the calibration\n  predictions without calling `AskUserQuestion` or writing the answers file.\n\n## Required capabilities & harness adapter\n\nThis workflow assumes a small set of capabilities. Map each one to your harness's actual tool. If a capability is missing, use the listed fallback.\n\n| Capability                  | Claude Code                          | OpenCode                                 | Codex CLI                                | Fallback if missing                                |\n|-----------------------------|--------------------------------------|------------------------------------------|------------------------------------------|----------------------------------------------------|\n| Read-only codebase research | `Agent` (subagent_type=Explore)      | `task` agent in read-only mode           | `apply_patch` read-only \u002F shell `grep`   | Do the search in-thread with `rg`\u002F`grep`           |\n| Foreground delegation       | `Agent` (subagent_type=general-purpose) | nested `task` agent, await output     | spawn nested CLI session                 | Do the work in-thread (more context cost)          |\n| Ask the user a question     | `AskUserQuestion`                    | inline prompt                            | inline prompt                            | Ask in the response and wait for the next turn     |\n| Subagent model tier         | Tier A: `opus` + `ultrathink`; Tier B: `sonnet` + medium thinking | Tier A: top `task` model + high reasoning; Tier B: smaller-tier `task` + medium reasoning | Tier A: top GPT + `reasoning_effort=high`; Tier B: current spark\u002Fcost-optimized Codex + `reasoning_effort=medium` | Honor the per-phase tier mapping in AGENTS.md; retry the same subagent UP on unavailability |\n\u003C!-- autospec-block:harness-adapter-core -->\n\n**Persistent project notes**: write durable preferences to **`AGENTS.md`** in the repo root — this is the de-facto standard recognized by Claude Code (also reads `CLAUDE.md`), OpenCode, and Codex. If your harness has its own private memory (e.g. Claude Code's `~\u002F.claude\u002F...\u002Fmemory\u002F`), mirror the same content there. Per AGENTS.md, subagent dispatches use a **two-tier policy**: Tier A (top model + extended thinking) for spec work and Tier B (cheaper model + medium thinking) for implementation work. Calibration interviewing is spec-tier judgment work: the answers file this skill writes is precedence-0 for every downstream persona-aware decision, so the repo-reading + question-generation + calibration-prediction agent runs on Tier A. The orchestrator keeps the user's invoked model. Fall back UP the tier on quota\u002Fcapacity or other unavailability by retrying the same subagent with the stronger tier while preserving parent context.\n\n---\n\n## Harness detection (run once at skill start, before reading the repo)\n\nDetect your harness by checking available tools before any phase:\n\n1. **Claude Code** — the `Agent` tool with a `subagent_type` parameter is available.\n   - `TIER_A` = `opus` + `ultrathink`  (the alias, which resolves to the current Claude Opus generation)\n   - `TIER_B` = `sonnet`               (the alias, which resolves to the current Claude Sonnet generation)\n\n2. **OpenCode** — a `task` tool with model\u002Ftier configuration is available (no `subagent_type`).\n   - `TIER_A` = top-tier task model + high reasoning\n   - `TIER_B` = smaller-tier task model + medium reasoning\n\n3. **Codex CLI** — neither `Agent` nor a configurable `task` tool is available; `apply_patch` is the primary edit tool.\n   - `TIER_A` = current top GPT model + `reasoning_effort=high`\n   - `TIER_B` = the spark \u002F cost-optimized variant of the configured model when one exists, else the configured model + `reasoning_effort=medium`\n\n**Fallback rule:** If `TIER_A` is not available in your harness (model unknown, quota\u002Fcapacity failure, authorization failure, or tool call returns an error for that model), silently retry the same subagent dispatch with the next available stronger model. Preserve the parent context on retry; for Codex native subagents, fork\u002Finherit the current conversation context and use the latest top GPT model instead of moving the work into the main session. Never ask the user.\n\nHold `TIER_A` and `TIER_B` for the entire skill run. Every \"Tier A\" and \"Tier B\" reference below resolves to these harness-specific values.\n\n## Answers-file contract\n\nThe interview reads and writes a single global file:\n\n```\n~\u002F.autospec\u002Foperator-persona.answers.json\n```\n\nShape (version 1):\n\n```json\n{\n  \"version\": 1,\n  \"batches\": [\n    {\n      \"id\": \"\u003Ctheme-slug>\",\n      \"status\": \"done\",\n      \"questions\": [\n        { \"q\": \"\u003Cquestion text>\", \"choice\": \"\u003Cselected option or null>\", \"free\": \"\u003Cfree-text or null>\" }\n      ]\n    }\n  ],\n  \"next_batch\": 1\n}\n```\n\n- `version` — always `1`; a file with a different `version` is treated as\n  incompatible and (with operator confirmation) reset.\n- `batches[]` — one entry per themed batch, in ask order. `status` is `done`\n  once every question in the batch is answered, otherwise `pending`.\n- `questions[]` — each carries the prompt `q`, the selected multiple-choice\n  `choice` (or `null` for free-text-only), and the `free` free-text note (or\n  `null`). A multiple-choice question has a non-null `choice`.\n- `next_batch` — the index into `batches[]` of the first batch whose `status` is\n  not `done`. When every batch is `done`, `next_batch` equals the batch count\n  (the terminal \"complete\" state).\n\nWrites are atomic (temp file + `mv`) so an interrupted batch never corrupts the\nfile.\n\n## Phase 1 — read the repo (Tier A)\n\n> **Model tier:** **TIER_A** (spec-tier judgment). Dispatch ONE read-only\n> research subagent to survey the repo and extract decision tensions; never\n> escalate to a second Tier-A pass for a single interview.\n\nDispatch a read-only research subagent to survey, in priority order:\n\n1. `docs\u002Fspecs\u002F*.md` — design tensions, deferred trade-offs, \"review\"-tagged\n   decisions.\n2. `docs\u002Fmemory\u002F` and the harness auto-memory — recurring operator preferences\n   and prior feedback lessons.\n3. `AGENTS.md` and any autonomy charter — standing rules and default-locks.\n4. Recent commits (`git log`) and open issues (`gh issue list`) — what the\n   operator actually prioritizes and rejects.\n5. The main code areas — the abstractions where judgment calls recur.\n\nThe subagent returns a ranked list of **decision tensions**: each a real,\nrecurring choice the operator makes (e.g. \"speed vs. correctness on flaky\ntests\", \"auto-merge vs. hold for review\"), with the option space for each.\n\n## Phase 2 — generate ≤50 questions in themed batches (hard cap)\n\nTurn the ranked tensions into questions, grouped into themed batches (e.g.\n\"merge autonomy\", \"test discipline\", \"spec rigor\"). Rules:\n\n- **Hard cap:** the total number of questions across all batches is **≤50**\n  (or `--max-questions N` if lower). If more than 50 candidate questions exist,\n  keep the 50 highest-ranked by recurrence × stakes; drop the rest. Never exceed\n  the cap — this is a closed invariant the test suite enforces.\n- **Multiple-choice preferred:** phrase each question as 2–4 concrete options\n  drawn from the repo's real option space. Use free-text only when the answer is\n  genuinely open-ended.\n- **Themed batches:** ask one `AskUserQuestion` call per theme (batch), so the\n  operator answers related questions together. Each batch becomes one\n  `batches[]` entry.\n\n## Phase 3 — interleave calibration questions\n\nWithin the batches, mark a subset of multiple-choice questions as **calibration\nquestions**: ones whose answer the agent can *infer* from memory\u002Fcharter\u002Fcommit\nhistory. For each, record the agent's **inferred prediction** before asking. The\noperator never sees the prediction; they just answer normally.\n\nAfter the interview, compute the **calibration-agreement %**:\n\n> agreement % = (number of multiple-choice calibration questions where the\n> operator's `choice` matched the inferred prediction) ÷ (total multiple-choice\n> calibration questions) × 100\n\nFree-text questions are **excluded** from the percentage (a free-text\ncalibration answer may optionally be LLM-graded as a side note, but never counts\ntoward the %). When the operator's choice **disagrees** with the prediction, the\ndisagreement **corrects the model** — update the per-dimension confidence and the\ninferred default for that tension, do not merely append the answer.\n\n## Phase 4 — resume logic\n\nOn every invocation (unless `--reset`):\n\n1. If `~\u002F.autospec\u002Foperator-persona.answers.json` is absent, start fresh with\n   `next_batch = 0`.\n2. If present and `version == 1`, read `next_batch` and **skip every batch whose\n   `status` is `done`**. Resume asking at `batches[next_batch]`. A `done` batch is\n   never re-asked.\n3. After each batch is fully answered, set its `status = done`, recompute\n   `next_batch` as the index of the first non-`done` batch (or the batch count if\n   none remain), and write the file atomically.\n4. When `next_batch` equals the batch count, the interview is complete; print the\n   summary and the calibration-agreement %.\n\n## Phase 5 — output\n\nOn completion, the answers file is the source-0 (global base) persona input\nconsumed downstream (F1\u002F#1417 reads it as precedence-0). Print:\n\n```\nautospec-persona summary\n- batches: \u003Cdone>\u002F\u003Ctotal>\n- questions asked: \u003CN> (cap: \u003Ccap>)\n- calibration (multiple-choice): \u003Cmatched>\u002F\u003Ctotal> = \u003Cagreement>%\n- answers file: ~\u002F.autospec\u002Foperator-persona.answers.json\n```\n\nPer-dimension confidence (raised on agreement, corrected on disagreement) is\nrecorded alongside each tension so downstream synthesis (F2) can weight it.\n\n## Hard rules\n\n- **≤50 questions, always.** The cap is a closed invariant; never ask more.\n- **Never re-ask a `done` batch.** Resume strictly at `next_batch`.\n- **Calibration % is multiple-choice only.** Free-text answers never count toward\n  the percentage.\n- **Disagreements correct the model**, they do not merely append a new answer.\n- **Atomic writes only** (temp + `mv`); an interrupted batch must not corrupt the\n  answers file.\n- This skill writes only `~\u002F.autospec\u002Foperator-persona.answers.json`; it never\n  edits repo files.\n","# autospec-persona (harness-neutral)\n\nStandalone calibration-interview skill. A Tier-A agent reads the current repo\n(`docs\u002Fspecs\u002F`, `docs\u002Fmemory\u002F`, `AGENTS.md`, the autonomy charter, recent\ncommits, open issues, and the main code areas), surfaces the real tensions and\nrecurring decisions, and turns them into **≤50 repo-grounded questions** (a hard\ncap) asked in themed `AskUserQuestion` batches (multiple-choice preferred). The\noperator's answers persist to `~\u002F.autospec\u002Foperator-persona.answers.json` and the\nrun is fully **resumable**: re-invoking continues at the next pending batch and\nnever re-asks a `done` batch.\n\nManage your own context — never exceed 60%. Delegate to subagents whenever your harness supports it.\n\n\u003C!-- autospec-block:startup-self-update SKILL_NAME=autospec-persona -->\n\n## Self-update mode\n\nIf the feature-request argument matches the regex `^\\s*update\\s*$` (case-insensitive, whitespace-padded), this skill enters self-update mode and does not run the normal interview:\n\n1. **Detect harness** by checking which install path exists for this skill:\n   - Claude Code: `~\u002F.claude\u002Fskills\u002Fautospec-persona\u002FSKILL.md`\n   - OpenCode:    `~\u002F.config\u002Fopencode\u002Fagent\u002Fautospec-persona.md`\n   - Codex CLI:   `~\u002F.codex\u002Fprompts\u002Fautospec-persona.md`\n2. **Re-install the full autospec suite from `main`** by piping the canonical installer:\n   ```bash\n   curl -fsSL https:\u002F\u002Fraw.githubusercontent.com\u002Fberlinguyinca\u002Fautospec\u002Fmain\u002Fbootstrap.sh | bash -s -- --skill all --harness all --update\n   ```\n   Run this one-liner once; it refreshes all autospec skills across all harnesses.\n3. **Show the diff** between the prior installed file(s) and the freshly fetched copy (e.g. `diff \u003C(cat \u003Cprior>) \u003C(curl -fsSL ...SKILL.md)` or the equivalent recorded by the installer).\n4. **Stop.** Do not read the repo, generate questions, or run the interview. Print the upgrade summary and return to the user.\n\nIf no install path is detected, print `Self-update: no installed copy of autospec-persona found; run install.sh first.` and exit.\n\n## Invocation\n\n```\n\u002Fautospec-persona [--reset] [--max-questions N] [--dry-run]\n```\n\n- `--reset` — discard any existing `~\u002F.autospec\u002Foperator-persona.answers.json` and\n  start a fresh interview from `next_batch=0`. Without it, the skill **resumes**.\n- `--max-questions N` — lower the question budget below the hard cap of 50\n  (never above; a value > 50 is clamped to 50).\n- `--dry-run` — generate and print the themed batches and the calibration\n  predictions without calling `AskUserQuestion` or writing the answers file.\n\n## Required capabilities & harness adapter\n\nThis workflow assumes a small set of capabilities. Map each one to your harness's actual tool. If a capability is missing, use the listed fallback.\n\n| Capability                  | Claude Code                          | OpenCode                                 | Codex CLI                                | Fallback if missing                                |\n|-----------------------------|--------------------------------------|------------------------------------------|------------------------------------------|----------------------------------------------------|\n| Read-only codebase research | `Agent` (subagent_type=Explore)      | `task` agent in read-only mode           | `apply_patch` read-only \u002F shell `grep`   | Do the search in-thread with `rg`\u002F`grep`           |\n| Foreground delegation       | `Agent` (subagent_type=general-purpose) | nested `task` agent, await output     | spawn nested CLI session                 | Do the work in-thread (more context cost)          |\n| Ask the user a question     | `AskUserQuestion`                    | inline prompt                            | inline prompt                            | Ask in the response and wait for the next turn     |\n| Subagent model tier         | Tier A: `opus` + `ultrathink`; Tier B: `sonnet` + medium thinking | Tier A: top `task` model + high reasoning; Tier B: smaller-tier `task` + medium reasoning | Tier A: top GPT + `reasoning_effort=high`; Tier B: current spark\u002Fcost-optimized Codex + `reasoning_effort=medium` | Honor the per-phase tier mapping in AGENTS.md; retry the same subagent UP on unavailability |\n\u003C!-- autospec-block:harness-adapter-core -->\n\n**Persistent project notes**: write durable preferences to **`AGENTS.md`** in the repo root — this is the de-facto standard recognized by Claude Code (also reads `CLAUDE.md`), OpenCode, and Codex. If your harness has its own private memory (e.g. Claude Code's `~\u002F.claude\u002F...\u002Fmemory\u002F`), mirror the same content there. Per AGENTS.md, subagent dispatches use a **two-tier policy**: Tier A (top model + extended thinking) for spec work and Tier B (cheaper model + medium thinking) for implementation work. Calibration interviewing is spec-tier judgment work: the answers file this skill writes is precedence-0 for every downstream persona-aware decision, so the repo-reading + question-generation + calibration-prediction agent runs on Tier A. The orchestrator keeps the user's invoked model. Fall back UP the tier on quota\u002Fcapacity or other unavailability by retrying the same subagent with the stronger tier while preserving parent context.\n\n---\n\n## Harness detection (run once at skill start, before reading the repo)\n\nDetect your harness by checking available tools before any phase:\n\n1. **Claude Code** — the `Agent` tool with a `subagent_type` parameter is available.\n   - `TIER_A` = `opus` + `ultrathink`  (the alias, which resolves to the current Claude Opus generation)\n   - `TIER_B` = `sonnet`               (the alias, which resolves to the current Claude Sonnet generation)\n\n2. **OpenCode** — a `task` tool with model\u002Ftier configuration is available (no `subagent_type`).\n   - `TIER_A` = top-tier task model + high reasoning\n   - `TIER_B` = smaller-tier task model + medium reasoning\n\n3. **Codex CLI** — neither `Agent` nor a configurable `task` tool is available; `apply_patch` is the primary edit tool.\n   - `TIER_A` = current top GPT model + `reasoning_effort=high`\n   - `TIER_B` = the spark \u002F cost-optimized variant of the configured model when one exists, else the configured model + `reasoning_effort=medium`\n\n**Fallback rule:** If `TIER_A` is not available in your harness (model unknown, quota\u002Fcapacity failure, authorization failure, or tool call returns an error for that model), silently retry the same subagent dispatch with the next available stronger model. Preserve the parent context on retry; for Codex native subagents, fork\u002Finherit the current conversation context and use the latest top GPT model instead of moving the work into the main session. Never ask the user.\n\nHold `TIER_A` and `TIER_B` for the entire skill run. Every \"Tier A\" and \"Tier B\" reference below resolves to these harness-specific values.\n\n## Answers-file contract\n\nThe interview reads and writes a single global file:\n\n```\n~\u002F.autospec\u002Foperator-persona.answers.json\n```\n\nShape (version 1):\n\n```json\n{\n  \"version\": 1,\n  \"batches\": [\n    {\n      \"id\": \"\u003Ctheme-slug>\",\n      \"status\": \"done\",\n      \"questions\": [\n        { \"q\": \"\u003Cquestion text>\", \"choice\": \"\u003Cselected option or null>\", \"free\": \"\u003Cfree-text or null>\" }\n      ]\n    }\n  ],\n  \"next_batch\": 1\n}\n```\n\n- `version` — always `1`; a file with a different `version` is treated as\n  incompatible and (with operator confirmation) reset.\n- `batches[]` — one entry per themed batch, in ask order. `status` is `done`\n  once every question in the batch is answered, otherwise `pending`.\n- `questions[]` — each carries the prompt `q`, the selected multiple-choice\n  `choice` (or `null` for free-text-only), and the `free` free-text note (or\n  `null`). A multiple-choice question has a non-null `choice`.\n- `next_batch` — the index into `batches[]` of the first batch whose `status` is\n  not `done`. When every batch is `done`, `next_batch` equals the batch count\n  (the terminal \"complete\" state).\n\nWrites are atomic (temp file + `mv`) so an interrupted batch never corrupts the\nfile.\n\n## Phase 1 — read the repo (Tier A)\n\n> **Model tier:** **TIER_A** (spec-tier judgment). Dispatch ONE read-only\n> research subagent to survey the repo and extract decision tensions; never\n> escalate to a second Tier-A pass for a single interview.\n\nDispatch a read-only research subagent to survey, in priority order:\n\n1. `docs\u002Fspecs\u002F*.md` — design tensions, deferred trade-offs, \"review\"-tagged\n   decisions.\n2. `docs\u002Fmemory\u002F` and the harness auto-memory — recurring operator preferences\n   and prior feedback lessons.\n3. `AGENTS.md` and any autonomy charter — standing rules and default-locks.\n4. Recent commits (`git log`) and open issues (`gh issue list`) — what the\n   operator actually prioritizes and rejects.\n5. The main code areas — the abstractions where judgment calls recur.\n\nThe subagent returns a ranked list of **decision tensions**: each a real,\nrecurring choice the operator makes (e.g. \"speed vs. correctness on flaky\ntests\", \"auto-merge vs. hold for review\"), with the option space for each.\n\n## Phase 2 — generate ≤50 questions in themed batches (hard cap)\n\nTurn the ranked tensions into questions, grouped into themed batches (e.g.\n\"merge autonomy\", \"test discipline\", \"spec rigor\"). Rules:\n\n- **Hard cap:** the total number of questions across all batches is **≤50**\n  (or `--max-questions N` if lower). If more than 50 candidate questions exist,\n  keep the 50 highest-ranked by recurrence × stakes; drop the rest. Never exceed\n  the cap — this is a closed invariant the test suite enforces.\n- **Multiple-choice preferred:** phrase each question as 2–4 concrete options\n  drawn from the repo's real option space. Use free-text only when the answer is\n  genuinely open-ended.\n- **Themed batches:** ask one `AskUserQuestion` call per theme (batch), so the\n  operator answers related questions together. Each batch becomes one\n  `batches[]` entry.\n\n## Phase 3 — interleave calibration questions\n\nWithin the batches, mark a subset of multiple-choice questions as **calibration\nquestions**: ones whose answer the agent can *infer* from memory\u002Fcharter\u002Fcommit\nhistory. For each, record the agent's **inferred prediction** before asking. The\noperator never sees the prediction; they just answer normally.\n\nAfter the interview, compute the **calibration-agreement %**:\n\n> agreement % = (number of multiple-choice calibration questions where the\n> operator's `choice` matched the inferred prediction) ÷ (total multiple-choice\n> calibration questions) × 100\n\nFree-text questions are **excluded** from the percentage (a free-text\ncalibration answer may optionally be LLM-graded as a side note, but never counts\ntoward the %). When the operator's choice **disagrees** with the prediction, the\ndisagreement **corrects the model** — update the per-dimension confidence and the\ninferred default for that tension, do not merely append the answer.\n\n## Phase 4 — resume logic\n\nOn every invocation (unless `--reset`):\n\n1. If `~\u002F.autospec\u002Foperator-persona.answers.json` is absent, start fresh with\n   `next_batch = 0`.\n2. If present and `version == 1`, read `next_batch` and **skip every batch whose\n   `status` is `done`**. Resume asking at `batches[next_batch]`. A `done` batch is\n   never re-asked.\n3. After each batch is fully answered, set its `status = done`, recompute\n   `next_batch` as the index of the first non-`done` batch (or the batch count if\n   none remain), and write the file atomically.\n4. When `next_batch` equals the batch count, the interview is complete; print the\n   summary and the calibration-agreement %.\n\n## Phase 5 — output\n\nOn completion, the answers file is the source-0 (global base) persona input\nconsumed downstream (F1\u002F#1417 reads it as precedence-0). Print:\n\n```\nautospec-persona summary\n- batches: \u003Cdone>\u002F\u003Ctotal>\n- questions asked: \u003CN> (cap: \u003Ccap>)\n- calibration (multiple-choice): \u003Cmatched>\u002F\u003Ctotal> = \u003Cagreement>%\n- answers file: ~\u002F.autospec\u002Foperator-persona.answers.json\n```\n\nPer-dimension confidence (raised on agreement, corrected on disagreement) is\nrecorded alongside each tension so downstream synthesis (F2) can weight it.\n\n## Hard rules\n\n- **≤50 questions, always.** The cap is a closed invariant; never ask more.\n- **Never re-ask a `done` batch.** Resume strictly at `next_batch`.\n- **Calibration % is multiple-choice only.** Free-text answers never count toward\n  the percentage.\n- **Disagreements correct the model**, they do not merely append a new answer.\n- **Atomic writes only** (temp + `mv`); an interrupted batch must not corrupt the\n  answers file.\n- This skill writes only `~\u002F.autospec\u002Foperator-persona.answers.json`; it never\n  edits repo files.\n",{"name":5,"description":7},"main","9e7f04357ff46ab731d4f56a4ac5ca5449c13bb0","skills\u002Fautospec-persona",1,"https:\u002F\u002Fgithub.com\u002Fberlinguyinca\u002Fautospec","https:\u002F\u002Fgithub.com\u002Fberlinguyinca","Love to lead, code, cook and contribute to open source and believe that you can learn anything you set your mind too.\r\n","2026-11-11 07:09:35",{"source":8,"skillName":5,"commands":58,"prompt":68,"zipAvailable":69},[59,62,65],{"tool":60,"command":61},"npx","npx skills add berlinguyinca\u002Fautospec --skill autospec-persona",{"tool":63,"command":64},"bunx","bunx skills add berlinguyinca\u002Fautospec --skill autospec-persona",{"tool":66,"command":67},"pnpm","pnpm dlx skills add berlinguyinca\u002Fautospec --skill autospec-persona","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fberlinguyinca\u002Fautospec\u002Fblob\u002Fmain\u002Fskills\u002Fautospec-persona\u002FSKILL.md",true,[71,77,83,87,91],{"path":72,"resourceType":73,"resourceTypeDesc":74,"sizeBytes":75,"sourceHtmlUrl":76},"codex\u002Fprompt.md","REFERENCE","参考文档",12639,"https:\u002F\u002Fgithub.com\u002Fberlinguyinca\u002Fautospec\u002Fblob\u002Fmain\u002Fskills\u002Fautospec-persona\u002Fcodex\u002Fprompt.md",{"path":78,"resourceType":79,"resourceTypeDesc":80,"sizeBytes":81,"sourceHtmlUrl":82},"install.sh","SCRIPT","脚本",9140,"https:\u002F\u002Fgithub.com\u002Fberlinguyinca\u002Fautospec\u002Fblob\u002Fmain\u002Fskills\u002Fautospec-persona\u002Finstall.sh",{"path":84,"resourceType":73,"resourceTypeDesc":74,"sizeBytes":85,"sourceHtmlUrl":86},"opencode\u002Fagent.md",13093,"https:\u002F\u002Fgithub.com\u002Fberlinguyinca\u002Fautospec\u002Fblob\u002Fmain\u002Fskills\u002Fautospec-persona\u002Fopencode\u002Fagent.md",{"path":88,"resourceType":73,"resourceTypeDesc":74,"sizeBytes":89,"sourceHtmlUrl":90},"README.md",3100,"https:\u002F\u002Fgithub.com\u002Fberlinguyinca\u002Fautospec\u002Fblob\u002Fmain\u002Fskills\u002Fautospec-persona\u002FREADME.md",{"path":92,"resourceType":79,"resourceTypeDesc":80,"sizeBytes":93,"sourceHtmlUrl":94},"uninstall.sh",3830,"https:\u002F\u002Fgithub.com\u002Fberlinguyinca\u002Fautospec\u002Fblob\u002Fmain\u002Fskills\u002Fautospec-persona\u002Funinstall.sh",[96,109,118,127,152,175],{"skillNo":97,"name":98,"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":99,"categories":100,"keywords":106,"publishedAt":44},"01M71RJHYMKPTC9AAPXWQ6F4RT","trio-pass-skill","https:\u002F\u002Fgithub.com\u002Fberlinguyinca\u002Fautospec\u002Fblob\u002Fmain\u002Ftests\u002Ffixtures\u002Flockstep-trio-pass\u002FSKILL.md",[101],{"categoryNo":102,"categoryName":103,"taxonomyType":104,"taxonomyTypeDesc":105},"01JBSEEDCAT000000000000023","Skill与智能体开发","SCENARIO","使用场景",[107,108],"trio","trio pass skill",{"skillNo":110,"name":111,"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":112,"categories":113,"keywords":115,"publishedAt":44},"01M71RJFBQ1KAF9XM3T3Y4VMSK","duo-pass-skill","https:\u002F\u002Fgithub.com\u002Fberlinguyinca\u002Fautospec\u002Fblob\u002Fmain\u002Ftests\u002Ffixtures\u002Flockstep-duo-pass\u002FSKILL.md",[114],{"categoryNo":102,"categoryName":103,"taxonomyType":104,"taxonomyTypeDesc":105},[116,117],"duo pass skill","duo",{"skillNo":119,"name":120,"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":121,"categories":122,"keywords":124,"publishedAt":44},"01M71RJDT1TB3P9TRWR6A534FB","duo-divergence-skill","https:\u002F\u002Fgithub.com\u002Fberlinguyinca\u002Fautospec\u002Fblob\u002Fmain\u002Ftests\u002Ffixtures\u002Flockstep-duo-divergence\u002FSKILL.md",[123],{"categoryNo":102,"categoryName":103,"taxonomyType":104,"taxonomyTypeDesc":105},[125,117,126],"duo divergence skill","divergence",{"skillNo":128,"name":25,"nameZh":6,"description":129,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":130,"categories":131,"keywords":135,"publishedAt":44},"01M71RJAC4NNEQSYBDSY3NZS73","Use when the user wants to ship a feature end-to-end across multiple commits — bootstraps repo if missing, brainstorms a design spec, decomposes into linked GitHub issues, or splits an existing spec into issues, then runs an autonomous implementation loop with auto-merge.","https:\u002F\u002Fgithub.com\u002Fberlinguyinca\u002Fautospec\u002Fblob\u002Fmain\u002Fskills\u002Fautospec\u002FSKILL.md",[132],{"categoryNo":133,"categoryName":134,"taxonomyType":104,"taxonomyTypeDesc":105},"01JBSEEDCAT000000000000015","视觉设计",[25,136,137,138,139,140,141,142,35,143,144,145,146,147,148,149,150,39,151,36],"ship a feature","end","issues","design","autonomous","auto","loop","across","design spec","merge","ship","implementation loop","github issues","feature","commits","missing",{"skillNo":153,"name":154,"nameZh":6,"description":155,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":156,"categories":157,"keywords":158,"publishedAt":44},"01M71RJ4F9HT8PT26CKNM9XE1M","autospec-upgrade","Use when the user wants to upgrade an Angular, Next.js, or React project to the latest official versions safely — by locking observable behavior before touching versions, upgrading one major at a time with official codemods, and gating completion on mutation score rather than coverage theater.","https:\u002F\u002Fgithub.com\u002Fberlinguyinca\u002Fautospec\u002Fblob\u002Fmain\u002Fskills\u002Fautospec-upgrade\u002FSKILL.md",[],[159,25,160,161,162,163,164,165,166,35,167,168,28,169,36,170,171,172,173,174],"autospec upgrade","upgrade","mutation","coverage","official","project","time","gating","angular","react","at","Next.js","one","with","behavior","latest",{"skillNo":176,"name":177,"nameZh":6,"description":178,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":179,"categories":180,"keywords":184,"publishedAt":44},"01M71RJ28FJ4PM6DXHB7T5CF8H","autospec-ui-audit","Use when a React Router application needs a deterministic whole-site route inventory before UI, UX, navigation, or evidence auditing begins.","https:\u002F\u002Fgithub.com\u002Fberlinguyinca\u002Fautospec\u002Fblob\u002Fmain\u002Fskills\u002Fautospec-ui-audit\u002FSKILL.md",[181],{"categoryNo":182,"categoryName":183,"taxonomyType":21,"taxonomyTypeDesc":22},"01JBSEEDCAT000000000000032","设计师",[185,25,186,187,188,36,189,190,191,192,193,194,195,196,197,168,198,35,199],"ui audit","audit","autospec ui audit","ui","inventory","evidence","deterministic","route","ux","site","react router","navigation","application","needs","router"]