[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M74ZMDH643C03HC71H6ZKPRT":3,"skill-same:01M74ZMDH643C03HC71H6ZKPRT":59},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":15,"categories":16,"keywords":17,"publishedAt":18,"linkOnly":19,"markdownContent":20,"markdownBody":21,"frontMatter":22,"refName":32,"commitSha":33,"skillDirectory":34,"forksCount":35,"repositoryHtmlUrl":36,"repositoryHomepageUrl":37,"ownerHtmlUrl":38,"ownerBlogUrl":6,"ownerDescription":6,"updatedAt":39,"install":40,"resources":52},"01M74ZMDH643C03HC71H6ZKPRT","sim-conductor",null,"Autonomously runs external user reaction simulations, internal audits, ideation scans, artifact validation, and quality reviews. Profiles the target artifact first, then derives task-appropriate personas, dispatches them as parallel agents, classifies findings into M\u002FS\u002FR tiers, and completes the pipeline through to commit automatically.","chrono-meta\u002Fforge-harness","chrono-meta","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F151837060?v=4","01M74ZDXK87ES4B30NKYSEH3MS","01M74ZDXM0F1NMYK7BS7PYR54G",33,"MIT","https:\u002F\u002Fgithub.com\u002Fchrono-meta\u002Fforge-harness\u002Fblob\u002Fmain\u002Fplugins\u002Ffh-meta\u002Fskills\u002Fsim-conductor\u002FSKILL.md",[],[],"2026-11-12 13:55:43",false,"---\nname: sim-conductor\ndescription: Autonomously runs external user reaction simulations, internal audits, ideation scans, artifact validation, and quality reviews. Profiles the target artifact first, then derives task-appropriate personas, dispatches them as parallel agents, classifies findings into M\u002FS\u002FR tiers, and completes the pipeline through to commit automatically.\nuser-invocable: true\nallowed-tools: [\"Read\", \"Write\", \"Bash\", \"Grep\", \"Glob\", \"Agent\"]\nmodel-note: session-inherit — Sonnet base is first-class (sonnet_floor_doctrine.md); depth-critical judged steps route to dispatch (opus agent \u002F cross-family sidecar, consent-gated), never a substrate requirement\n---\n\n# sim-conductor — Meta-Simulation Automation Orchestrator\n\n> Profiles target → derives personas → dispatches parallel agents → M\u002FS\u002FR triage → commit.\n> Personas are sourced installed-first → **filled from the persona-container schema** (7 slots as of 2026-07-18 — stance added, not an\n> ad-hoc directive) → external install. Scale: 3 to 16, extensible to a crowd behind the\n> marginal-coverage stop (§Scale).\n\n## Invocation Triggers\n\n| Trigger pattern | Suggested Area |\n|---|---|\n| \"Look at this from an external user's perspective\", \"How would someone else see this?\" | Area A |\n| \"Check the harness\", \"asset audit\", \"internal audit\" | Area B |\n| \"Any new ideas?\", \"ideation scan\", \"naming candidates\" | Area C |\n| \"Review this code\" (pre-PR), \"review this prompt\" | Area D (code) |\n| \"Check if the session card is correct\", \"does the skill actually work\", \"memory cold-start validation\" | Area D (consumer) |\n| \"Is the output good?\", \"artifact quality review\" | Area E |\n\nProposal format: `\"If it's related to [X], should I simulate with \u002Fsim-conductor [Area]?\"`\n\n### Natural Language Triggers\n\n| Example utterance | Intent | Mapped Area |\n|---|---|---|\n| \"Validate our system virtually\" | Full system simulation | Area B |\n| \"Test it with personas\", \"Run an external user simulation\" | External user reaction | Area A |\n| \"Look at it through someone else's eyes\" | External perspective | Area A |\n| \"Validate that this actually works\" | Real-usage validation | Area D |\n| \"Find problems aggressively\" | Adversarial validation | Area B (`challenger`) |\n\n## Triggers\n\n```\n\u002Fsim-conductor                        # default: Area C auto-run\n\u002Fsim-conductor A                      # external user simulation\n\u002Fsim-conductor B                      # internal meta audit\n\u002Fsim-conductor C                      # Innovator scan\n\u002Fsim-conductor all                    # full A + B + C\n\u002Fsim-conductor A --target \u003Cpath>      # specific asset target\n\u002Fsim-conductor D --target \u003Cfile>      # code\u002Fprompt multi-persona review (pre-PR)\n\u002Fsim-conductor D session [--target]   # session card cold-start validation\n\u002Fsim-conductor D skill \u003Cname>         # skill trigger utterance consumer simulation\n\u002Fsim-conductor D memory \u003Cfile>        # memory file cold-start validation\n\u002Fsim-conductor E --target \u003Cfile>      # artifact quality review (post-pipeline)\n```\n\n**Frequency limits**: Area A after asset changes or manual trigger · Area B **min 7-day interval** (executor-applied guard: stop a re-run inside 7 days) · **event-bound, no periodic cadence** (operator decision 2026-10-19): run Area B before a minor\u002Fmajor release or a public announcement — the moment an outside reader's view pays — unless one already ran since the previous minor release. The trigger lives in `templates\u002FPRE-PUBLISH-CHECKLIST.md §Read it as a reader`; `fh_audit_check.zsh` no longer nags on a 30-day clock · Area C\u002FD\u002FE no limit. Why: per-change 4-axis gates with a cross-family reviewer, the nightly live-eval and the weekly harvest cover what a monthly whole-plugin pass used to; what only Area B sees (whole-set naming gaps, an outsider's first read) matters at publication.\n\n---\n\n## Step 0 — Environment Sync\n\nAuto-detect harness root, sync sim clone if present, set `$REPORT_DIR`.\n\n> **Detail**: See `SKILL_detail.md §Step0-Bash` — harness root detection, sim clone sync, report dir creation — read when executing Step 0.\n\n**Area B frequency check** (the 7-day *minimum* interval): Before running Area B, inspect the latest Area B report in `$REPORT_DIR`; if it is less than 7 days old, stop and report its date. This is an executor-applied procedure, not an automatic shell gate. No report in the selected `$REPORT_DIR` → treat as first run for this check. There is no maximum-interval reminder any more — when to run is decided by the release\u002Fannouncement event above.\n\n---\n\n## Step 0.3 — Target Profile Analysis\n\n> **Schema**: `knowledge\u002Fshared\u002Fharness-core\u002Ftpa_schema.md` — canonical field definitions, artifact type table, gate routing. Use §Profile Output Format for output block.\n\nRuns when Area is unspecified, or when target file is provided before Area selection.\n**Skip** if user provides a complete directive (e.g. `\u002Fsim-conductor A --target \u003Cfile>`).\n\nRead target artifact(s) → classify on 5 dimensions → output recommendation → user confirms or overrides:\n\n| Dimension | Signal → Weight shift |\n|---|---|\n| `artifact_type` | SKILL.md \u002F design-doc → Area B + D-skill↑ · README \u002F CHEATSHEET → Area A↑ · code \u002F config → Area D-code↑ |\n| `audience` | external installer \u002F first-time user → beginner↑ · internal team only → challenger↑ |\n| `claim_density` | 3+ stated benefits or superlatives → challenger↑ |\n| `risk_level` | external publish \u002F marketplace listing → steel-quench prerequisite triggered. **Mechanical floor (not judge-only)**: any of — publish\u002Fmarketplace target · public-surface or visibility change · auth\u002Fsecret-handling or executable code · an FH asset under the 4-axis gate — **forces `risk_level ≥ medium`** regardless of profiler judgment. The floor closes the \"fool the profiler into `low` to skip Step 0.6\" seam; the judge may only raise above the floor, never below it. |\n| `novelty` | first-of-its-kind \u002F no prior session evidence → phantom-quench recommended |\n\n```\nTarget Profile output:\n  artifact_type: [type]\n  audience: [internal | external | mixed]\n  claim_density: [low | medium | high] ([N] stated claims)\n  risk_level: [low | medium | high]\n\nRecommendation:\n  Areas: [list + rationale]\n  Persona composition: [list + weight]\n  Scale: [Minimum 3 | Extended 4–8 | Full ≤16]\n  Prerequisites: [steel-quench \u002F phantom-quench \u002F none]\n```\n\n#### Persona Discovery (after profile → before dispatch)\n\n> **Schema**: `knowledge\u002Fshared\u002Fharness-core\u002Fpersona_container_schema.md` — the 7-slot persona-definition\n> container. ② below **fills** these slots (lens \u002F internal-logic \u002F external-grounding \u002F output-protocol \u002F stance\n> \u002F cost-tier \u002F lifecycle) instead of injecting a shapeless directive. The container is what makes a\n> synthesized persona reproducible and cost-routable; the old \"fallback palette\" was undefined.\n\nAfter profiling, sim-conductor determines the needed perspective types, then maps each to the best available agent:\n\n```\nFor each needed perspective:\n  ① Scan installed plugins + .claude\u002Fagents\u002F → exact match? → use it\n  ② Fill the 7-slot persona container — lens · internal-logic · external-grounding · output-protocol · stance\n       (= Step 1.5 parallax shape) · cost-tier · lifecycle → a defined, reproducible persona; the filled\n       container IS the dispatch prompt (replaces the old ad-hoc \"fallback palette\" directive). Assign\n       the cost-tier slot per §Scale cost-tier routing. [Schema: persona_container_schema.md]\n  ③ GAP: no ①② match for a high-weight perspective\n       → query \u002Fplugin-recommender: \"find agents for [perspective] matching [artifact_type] context\"\n       → present: name · install command · estimated fit · token cost\n       → user chooses: install now \u002F skip \u002F substitute with ②\n```\n\nPersona Discovery output:\n\n```\nPersona Map:\n  beginner       → [installed agent: beginner] OR [filled 7-slot container]\n  challenger     → [installed agent: challenger] OR [filled 7-slot container]\n  [profile-specific role] → ⚠️ GAP — plugin-recommender recommends: [X] (install? y\u002Fn)\n```\n\nGaps on **high-weight personas** (from profile) block dispatch until resolved. Gaps on low-weight personas → auto-fill with built-in fallback, no block.\n\n**Degraded coverage rule**: When user skips a high-weight gap (substitutes with built-in fallback), flag in the Step 3 report under \"Persona composition used\" as `⚠️ degraded: [perspective]`. Do not silently proceed. Degraded coverage on risk_level=high targets → additionally warn before dispatch.\n\nOverriding a prerequisite recommendation requires explicit \"skip [prerequisite]\" utterance.\n\n> **Detail**: See `SKILL_detail.md §PersonaDiscovery` — plugin-recommender query format, gap resolution decision tree, persona map examples for each artifact type — read when running Persona Discovery.\n\n> **Detail**: See `SKILL_detail.md §Profile-Examples` — worked profiles for SKILL.md, README, bash script, design doc — read when result is ambiguous or to calibrate weights.\n\n---\n\n## Step 0.5 — Precondition Validation\n\nProactively surface 1 concern, then confirm whether to proceed. Skip if no concern.\n\n| Check | Concern trigger condition |\n|---|---|\n| Area B unresolved S-tier | An unresolved S-tier finding from the previous run |\n| Unresolved M-tier | New simulation without processing previous M-tier |\n| steel-quench not run | Area A in external publish context → **mandatory gate: Area A does not proceed until steel-quench completes** |\n| Self-recursive path | sim-conductor attempting to auto-PR its own M-tier |\n| Goal-method alignment | Requested Area doesn't match actual problem |\n| D\u002FE --target missing | Area D(code)\u002FE runs without --target file |\n| D session target not found | Session start card file not found → confirm path |\n| E execution timing error | Area E before pipeline run (no artifact yet) |\n\nConcern format: `\"One thing to check before [Area X]: [concern]. Proceed?\"`\n\n---\n\n## Step 0.6 — Cross-Model Coverage Gate (risk≥medium — hard)\n\nCloses the homogeneous-blind-spot + formatting-flattery vector (judge-robustness swarm, 2026-06-13):\na panel of same-session Claude sub-agents shares one model's blind spots, so a clean verdict can be\nflattery the **whole panel** is blind to — and with no quotable-rule violation, no persona escalates.\nFor `risk_level ≥ medium` targets (from Step 0.3), at least one persona MUST come from a source\n**outside the orchestrator's own session context**. This **promotes the former advisory \"dual\nvalidation principle\"** (detail §AreaB-Baseline #4) to a hard gate — the mechanical-anchor pattern of\nhardening #1–#5: a judged verdict binds to a fact the judge cannot fake.\n\n**Graceful-degradation ladder** — take the highest available rung. The gate **never breaks\nsim-conductor**; at the bottom it only withdraws the *unsafe autonomy* (self-certifying a blind verdict),\nnot the run:\n\n| Rung | Source | `cross_model_coverage` | Closes | When |\n|---|---|:---:|---|---|\n| 1 | External CLI team (Multi-Team Mode — §MultiTeam) | `external` | model-level blind spot (genuine cross-model) | 1+ external CLI live + probe non-empty |\n| 2 | Cross-session Claude — `claude -p` headless, or an Agent with **zero inherited context** | `cross-session` | **session-contamination only** — a fresh Claude shares the same weights\u002FRLHF gradient, so it does **not** close the model-level blind spot; it only removes the orchestrator's working-memory bias. Honest partial mitigation, labeled as such | no external CLI; dispatch probe returns non-empty |\n| 3 | Same-session sub-agents only | `NONE` | nothing — homogeneous panel | neither rung's probe succeeded |\n\n**Mechanical anchor** — `cross_model_coverage` is valid **only if backed by a quoted dispatch artifact**,\nnot a self-assessment (the self-signing hole hardening #1 closed for the marker — same fix here). To\nrecord `external` or `cross-session`, the Step 3 report must **quote a non-empty excerpt of the actual\ndispatch output** (external CLI stdout, or the dispatched Agent's returned verdict text); a label with no\nquoted excerpt is invalid and falls to `NONE`. **Liveness, not mere availability**: probe the rung before\nclaiming it — attempt the dispatch with a timeout; if it errors or returns empty (plan-gate closed,\ncontext saturated, CLI present-but-dead), record `NONE`, never assume the rung succeeded. This is the\nsame honest scope as #1: the artifact makes the claim **auditable**, not cryptographically unforgeable —\na fabricated excerpt remains the operator's + weekly-audit's residual by design. On rung 3 (`NONE`) for a\nrisk≥medium target:\n- the report flags `⚠️ cross-model coverage: NONE — homogeneous same-session panel; verdict provisional`, **and**\n- **Step 4 auto-commit privilege is withdrawn** (see Step 4): M-tier fixes may be *prepared* but the\n  commit waits for the operator's explicit go. Auto-committing a structurally self-blind verdict is\n  exactly what the exploit targeted — so that single privilege is what degrades, not the simulation.\n\n`risk_level = low` targets are exempt (a homogeneous panel is acceptable); the gate fires only at\nmedium+. The rung-2 fallback is what makes this CC-only-safe: a Claude-only environment still gets a\nreal cross-context read (a fresh isolated dispatch shares no working memory with the orchestrator), so\n`NONE` is reached only when *both* external CLIs and a second Claude context are unavailable.\n\n---\n\n## Step 1 — Area-Specific Simulation\n\n### Area A — External User Perspective\n\nTarget: skill descriptions, README, CHEATSHEET entry points.\n\n#### Task-Adaptive Persona Selection\n\nsim-conductor does **not** run a fixed persona set. It derives needed perspectives from the task, then sources each persona in priority order:\n\n```\n① Installed first — scan installed plugins + .claude\u002Fagents\u002F for a matching persona\u002Fagent\n② Container fill — fill the 7-slot persona container for the standpoint; the filled container IS the\n     dispatch prompt into a general-purpose Agent (Path B — replaces the old ad-hoc role directive)\n③ External fetch — chain to \u002Fplugin-recommender when ①② insufficient for high-stakes tasks\n```\n\nShipped standpoint agents (① tier — sourced installed-first):\n\nFH ships a coherent **user-mastery spectrum** as real, reusable agents (not prompt-directive shells) — found by the ① installed-first scan, reusable across skills, and each isolated-context dispatchable for a true cold read (the bias-isolation value: an evaluator outside the author's context reads cold):\n\n| Agent | Spectrum tier | Standpoint | Type |\n|---|---|---|---|\n| `beginner` | entry | First-contact cold-read — onboarding friction a fluent author cannot feel | reasoning |\n| `main-player` | core | Engaged user; intelligently scopes Light \u002F Midcore \u002F Heavy (Heavy = classic power-user edge\u002Flimit lens) | reasoning |\n| `expert` | frontier | Domain authority; web-grounded accuracy + SOTA currency, citation-enforced | data (WebSearch\u002FWebFetch) |\n| `challenger` | adversarial axis | Frontier adversary; U1 absorbs the skeptic \"why not just X?\" lens | adversarial |\n\n> **Lineage**: `beginner` \u002F `main-player` \u002F `expert` are the FH-native frontier successors to the field deep-insight `user` group (newcomer \u002F power-user) — re-derived to FH grade with embedded methodology + Done-When, not name-copied. `challenger` is the advanced form of the field `devil-advocate`. The former standalone skeptic standpoint is folded into `challenger` U1.\n\n**Ad-hoc roles** (② tier — container fill): when the profile demands a standpoint with no shipped agent (e.g. \"security auditor\", \"non-native reader\"), fill the 7-slot persona container for that standpoint — the filled container is the dispatch prompt. Prefer ① shipped agents; use ② only for genuinely task-specific one-offs.\n\n#### Scale\n\n| Scale | Count | When |\n|---|---|---|\n| **Minimum** | 3 | Routine Area A — most task-relevant 3 perspectives |\n| **Extended** | 4–8 | High-stakes publish \u002F external release |\n| **Full** | Up to 16 parallel | Pre-major-version \u002F architecture review |\n| **Crowd** | 16+ — **no fixed cap; bounded by the marginal-coverage stop** | Only when the marginal-coverage stop has demonstrably **not** fired at 16 — i.e. persona #17 is still decorrelated and finding-productive. **Not** chosen by artifact size or audience count (Full already covers those); the stop, not a headcount, opens this tier |\n\n**Crowd is output-driven, never a target** (per the container schema): the 16-cap lifts only behind the\n**marginal-coverage stop** — add a persona only if it is *decorrelated* (a distinct viewpoint, or — at\ncrowd scale — a different model family) from those already running; **stop when the marginal new-finding\nrate → 0**. A crowd of identical-output personas is decorative (steel-quench Wave-1 angle #1). Do not\nname a headcount goal; the count is whatever the stop allows.\n\n**Cost-tier routing (the economy lever)** — each persona's `cost-tier` slot routes it to a model so\nbreadth is cheap and the verdict stays trustworthy:\n- **floor-local \u002F heavy-local** (e.g. a local Ollama sidecar) carry cheap *breadth* — extra decorrelated\n  standpoints at near-zero token cost. A heavier local (27–32B) is a stronger canary than a floor model,\n  but **on one probe, not a benchmark** (`[[persona-container-schema]]` §tier-distribution) — so a local\n  tier is a **canary, never the gate**.\n- **frontier** carries the **terminal verdict** — the M\u002FS\u002FR triage and any auto-commit decision stay\n  frontier. This is the same anchor as Step 0.6: local breadth *decorrelates*, it does not *decide*.\n- Cost-tier routing and the Step 0.6 cross-model gate are complementary, not duplicate: 0.6 *requires* a\n  non-session source for risk≥medium (blind-spot closure); cost-tier *distributes* breadth for economy.\n  A local-sidecar persona can satisfy both at once (cross-family **and** cheap), but only Step 0.6's\n  quoted-artifact rule lets it count as `external` coverage.\n\nAll personas run as **parallel Agents** — no sequential bottleneck. Use agent-composer Medium\u002FLarge\nfan-out for Extended\u002FFull; Crowd scale dispatches in decorrelated waves, re-checking the marginal-coverage\nstop between waves rather than firing all at once.\n\n**Simplification guard**: Routine internal audits → ② built-in fallback at Minimum scale. Chain to ③ only when a needed perspective has no ①② match, or stakes are high.\n\n#### Multi-Team Mode (when external CLIs available)\n\nWhen 1+ external CLIs detected, Area A upgrades to Multi-Team Panel: each CLI forms an independent team. Cross-team: 2+ teams flag same issue → tier-up (S→S-confirmed). External-only findings → \"Claude blind spot\" flag.\n\nPre-entry user confirmation required before multi-team execution.\n\n> **Detail**: See `SKILL_detail.md §MultiTeam` — team formation table (T0–T4), CLI detection bash, confirmation dialog, cross-team synthesis format — read when multi-team mode activates.\n\n**A-1** (`beginner`) — first-contact friendliness · onboarding friction · terminology clarity\n**A-2** (`main-player`) — engaged-use fit; Heavy tier: install conflicts · duplication · silent overwrite\n**A-3** (`challenger`, artifact_type=\"SKILL\") — claim-evidence gaps · angles U3, U5, S2\n\n> ⚠️ **Human review gate**: Area A S-tier judgments require owner review before entering AI-AI loop.\n\n---\n\n### Area B — Internal Meta Audit\n\nTarget: all fh-meta assets (skills + agents + plugin.json).\n\n> **Self-recursive caution**: When M-tier found in sim-conductor itself → report only, delegate to user.\n\n**Dispatch pattern**: hub-persona-auditor ∥ persona-innovator (parallel) → challenger (after, with both outputs as context)\n\n1. **hub-persona-auditor** — README\u002FCHEATSHEET as external briefing, 4-axis review, 3-tier suggestions\n2. **persona-innovator** (Mode I) — naming gaps + structural gaps + 3-5 candidates\n3. **challenger** (artifact_type=\"SKILL\") — receives 1+2 outputs, angles U2, D3, U5 — \"what was missed?\"\n\nRationale: 1+2 run without influencing each other; challenger synthesizes both perspectives independently, avoiding Cost of Consensus.\n\n> **Detail**: See `SKILL_detail.md §AreaB-Baseline` — external baseline injection methods, dual-validation principle, Area B → steel-quench handoff — read when setting up Area B defenses.\n\n> ⚠️ **Human review gate**: Area B convergence is provisional until owner confirms.\n\n---\n\n### Area C — Innovator Scan\n\nRuns independently. Skip if included in Area B.\n\n**persona-innovator** (Mode F = Internal + External): current harness gaps + frontier scan. Output: naming candidates + absorption signals + top 1 priority action.\n\n---\n\n### Area D — Artifact Validity Validation\n\n| Mode | Trigger | Core question |\n|---|---|---|\n| **code** | `\u002Fsim-conductor D --target \u003Cfile>` | Is the code\u002Fprompt well-built? |\n| **consumer** | `\u002Fsim-conductor D session\u002Fskill\u002Fmemory` | Does the artifact work for the intended consumer? |\n\n#### D-code — Profile-Aware Persona Routing\n\nPersona composition adapts to `artifact_type` from Step 0.3 profile:\n\n| Artifact type | Primary persona | Supporting persona | Focus |\n|---|---|---|---|\n| SKILL.md \u002F design doc | challenger (artifact_type=\"SKILL\") | `beginner` | Governance gaps, behavioral rule coverage |\n| Python \u002F JS \u002F bash code | challenger (artifact_type=\"Code\") | `main-player` (Heavy) | Edge cases, performance, security surface |\n| Prompt \u002F config | `beginner` | challenger | Interpretation errors, implicit assumptions |\n| Auth \u002F security-sensitive | challenger + Security-auditor† | `main-player` (Heavy) | Attack surface, privilege escalation |\n\n† Security-auditor = ② tier — the 7-slot persona container filled for a security-auditor standpoint.\n\nAll personas run in parallel. Findings = M\u002FS\u002FR → M-tier items fixed immediately.\n\n#### D-consumer — Artifact Consumer Simulation\n\nConsumer agent attempts actual use (not just reads and judges). Grades: F (functional) \u002F P (partial → S-tier) \u002F B (broken → M-tier immediate fix).\n\n> **Detail**: See `SKILL_detail.md §AreaD-Consumer` — D-session 4 questions, D-skill cold-start procedure, D-memory check items, verdict criteria — read when executing D-consumer.\n\n---\n\n### Area E — Artifact Quality Review\n\n`expert` objection (E-1) + Practitioner confusion (E-2) in parallel → Pattern structuring (E-3) integrates both.\n\n> **Detail**: See `SKILL_detail.md §AreaE-Detail` — E-1\u002FE-2\u002FE-3 execution, finding format, pattern naming procedure — read when executing Area E.\n\n---\n\n## Step 1.5 — Persona Output Protocol + Neutral Synthesizer (parallax)\n\nGeneralized from the field `deep-insight` multi-persona pattern (private companion store), domain-stripped — the *pattern*\nis renamed **parallax** for public FH (it is a mode of this skill, not a separate skill — see asset-placement\n2026-06-06). It gives the persona dispatch above a shared output contract + a neutral aggregator, so\nmulti-persona findings stay comparable and the synthesis injects no bias of its own.\n\n> **Naming provenance (precise)**: \"renamed\" above refers to the *pattern* (→ parallax), not the personas.\n> The company-team-coupled field personas (fe\u002Fbe\u002Fios\u002Fpm\u002Fux-writer\u002Fcompliance\u002Fqa) were domain-stripped\n> entirely. The generic `user` group (newcomer\u002Fpower-user) was **re-derived to FH grade as the shipped\n> `beginner` \u002F `main-player` \u002F `expert` mastery-spectrum agents** (embedded methodology + Done-When, not\n> name-copied shells), and the field `devil-advocate` was **advanced into `challenger`** (sandboxed-adversary\n> + adaptive attack matrix). Lineage is acknowledged; nothing is carried verbatim as a shell.\n\n**Shared persona output protocol** — every dispatched persona emits the same shape, whatever its lens:\n\n```\n### Strengths        (0–3, from this persona's viewpoint)\n### Concerns\n  Critical   — compile\u002Fruntime failure · clear logic error · data corruption · security leak\n  Important  — significant user\u002Fservice impact in a plausible scenario\n  Suggestion — optional improvement\n  (each item: [file:line or quoted span] one-line summary — rationale)\n### Open questions   (0–3 items needed for a decision)\n### Absence check    (outside-vantage personas — beginner\u002Fintegrator: what does the artifact FAIL to\n                      specify that this standpoint needs? discoverability · undocumented contract ·\n                      unstated assumption. A normal, self-administrable rubric item — surfaces real gaps.)\n```\n\n**FP judgment discipline** — only escalate when confident. Never escalate: pre-existing issues not\nintroduced by this change · style\u002Fquality without a quotable rule · linter-catchable · speculative\n(\"might break if…\") · subjective preference (→ Suggestion). **If not confident, do not mark it** — false\npositives erode reviewer trust.\n\n**Neutral synthesizer** — the aggregator is a NON-persona; it adds no opinion of its own:\n- No opinion injection — never a conclusion no persona stated.\n- Preserve attribution — always traceable which persona said what.\n- Priority labels verbatim — Critical\u002FImportant\u002FSuggestion carried as the persona set them.\n- No forced consensus or forced conflict — report Common opinions (2+ personas agree) and Conflicts\n  (position A vs B, each with rationale) as-is. Feeds Step 2 M\u002FS\u002FR triage (M ← Critical or 2+ personas).\n- Cost-tier aware triage — a frontier-tier persona's verdict is terminal; a floor\u002Fheavy-local persona's\n  finding enters only as decorrelated **breadth** (it can raise a standpoint to `covered`, never solely\n  set an M-tier on its own). This enforces in the aggregator the §Scale \"local breadth decorrelates, it\n  does not decide\" lock — without it, a local canary's finding would drive triage identically to frontier.\n\n**Zero-coverage map (mandatory synthesizer output)** — the synthesizer must emit, mechanically, the set\nof standpoints that produced **no** finding, not only the ones that did (judge-robustness swarm,\n2026-06-13). Enumerate the persona **standpoints** in play — those the Step 0.3 profile recommended,\n**plus the standpoints its dimensions imply** (risk_level=high → a security\u002Fpublish standpoint ·\naudience=mixed → a non-native-reader standpoint · claim_density=high → a claim-evidence standpoint).\nList the *standpoints*, not the bare dimension names (a `risk_level (low) → ZERO` row is noise that\ntrains operators to ignore ⚠️). Mark each `covered` (≥1 persona addressed it) or `ZERO` (no persona\ntouched it):\n\n```\nCoverage map:\n  beginner (onboarding friction)      → covered (A-1: 2 findings)\n  challenger (claim-evidence)         → covered (A-3: 1 finding)\n  security surface (risk_level=high)  → ZERO ⚠️\n  non-native reader (audience=mixed)  → ZERO ⚠️\n```\n\nA clean report with `ZERO` standpoints is **not** a pass — it is an uncovered surface, reported as such.\nThis converts the formatting-flattery failure (everything reads fine → nothing escalates) into a visible\ngap: **silence on a standpoint is reported as `ZERO`, never inferred as approval.** It is a checklist\nderived from the profile and the dispatch outputs — a mechanical anchor, not a judgment. Carry the map\nverbatim into the Step 3 report.\n\nThe two severity vocabularies are layered, not redundant: a persona running **in isolation** assigns only\nits own Critical\u002FImportant\u002FSuggestion — it cannot assign M\u002FS\u002FR, since `S = found by 3+ personas` depends on\ncross-persona agreement the isolated persona never sees. The synthesizer is the only context that can triage\nto M\u002FS\u002FR. So isolation *requires* the per-persona → synthesized two-layer split.\n\n**External-harness persona sourcing** — isomorphic to steel-quench Step 0.4 (Specialized Reviewer\nDiscovery) + Wave 5 (external CLI teams). A needed lens may be sourced from an **installed sibling\nharness**, not only the built-in palette — e.g. gstack `\u002Freview` (staff-engineer), `\u002Fcso`\n(security-officer), `\u002Fqa` (QA-lead) when gstack is installed. sim-conductor orchestrates; the sibling\nsupplies the specialist lens. Same ①installed → ②fallback → ③fetch priority as Persona Discovery — an\nexternal harness's review-skills count as ① installed sources, widening the persona pool without FH\nshipping every specialist.\n\n> **Absence check — resolved (added above)**: the clean replication (companion-store experiment, Arm F: identical\n> artifact, explicit omission prompt, ownership-only variable) found self ≈ isolated (~90% overlap) —\n> **omission-detection is self-administrable when explicitly asked**, refuting the earlier \"the author\n> can't see their own omissions\" (Arm E, a prompt\u002Fdesign-drift artifact). So the Absence check is a\n> normal, valuable rubric item (it surfaces real undocumented-contract gaps), not a special isolation-only\n> power. The pattern's value is rubric\u002Fstandpoint *supply + routine enforcement* (copyable utility), not\n> de-biasing — see `knowledge\u002Fshared\u002Fpatterns\u002Fmulti-persona-review.md`.\n\n---\n\n## Step 2 — Synthesis\n\n| Tier | Criteria | Action |\n|---|---|---|\n| **M (Mandatory)** | Blocks external user entry \u002F naming-substance mismatch \u002F security\u002Ftrust damage | Immediate fix required |\n| **S (Strong)** | Feature degradation \u002F found by 3+ personas | Address within next session |\n| **R (Recommended)** | Improvement value \u002F found by single persona | Backlog |\n\n**Deduplication**: Multiple personas → same location: M > S > R.\n\n**Early Convergence Detection**: 2+ Areas independently classify same item as M-tier → immediate fix, remaining Areas continue for S\u002FR coverage only.\n\n---\n\n## Step 3 — Report\n\nFile: `$REPORT_DIR\u002Fsim_YYYY_MM_DD_area_[X].md`\n\n> **Detail**: See `SKILL_detail.md §Report-Format` — full report template with frontmatter, M\u002FS\u002FR sections, emerging asset candidates — read when writing the report.\n\n---\n\n## Step 4 — PR or Direct Commit\n\n1+ M-tier → fix immediately → commit. PR creation requires explicit user request.\n0 M-tier → commit report only + report S\u002FR backlog.\n\n**Cross-model gate on auto-commit (risk≥medium)** — when Step 0.6 recorded `cross_model_coverage: NONE`\non a risk≥medium target, the auto-commit privilege is **withdrawn**: prepare the M-tier fixes and write\nthe report, but do **not** self-commit — surface *\"cross-model coverage NONE on a risk≥medium target;\nthe verdict is from a homogeneous same-session panel. Commit the fixes, or add a cross-context read\nfirst?\"* and wait for the operator's go. `external`\u002F`cross-session` coverage, or risk_level=low, commits\nas normal. (This withdraws one privilege, not the run — the report and fixes still exist.)\n\n> **Detail**: See `SKILL_detail.md §PR-Bash` — branch creation bash, commit + push, gh pr create template — read when creating a PR.\n\n---\n\n## Step 5 — Graduate a Proven Persona (optional, rare)\n\nMost synthesized personas are **throwaway** — the run accelerated the work and the persona is discarded.\nA persona earns a permanent home only by passing the container schema's **admission test** (the analogue\nof \"app-ification\"; `[[persona-container-schema]]` §Lifecycle):\n\n| Check | Criterion | Class |\n|---|---|---|\n| **Recurrence** | surfaced **≥1 unique S\u002FM finding** (one no other persona in the run produced) across **≥2 distinct targets** — not a one-off lucky fire | measured |\n| **Trust bar (skills only)** | if graduating into a *skill* rather than a project's persona set, additionally clears the **FH 6-item creation gate** | mandatory-pass |\n\n- **Pass → into a project's static set**: route the write through `asset-placement-gate` (expected\n  verdict: project-local) before writing the filled container to the project's persona file (e.g.\n  the-bible `core\u002Fpersonas.json`) — sim-conductor proposes the target, the placement gate confirms it,\n  the operator approves. A project-local asset, not an FH change, but the cross-boundary write is still gated.\n- **Pass → into a skill**: route through `asset-placement-gate` + the New-Skill creation gate (the\n  skill-foundry inversion — the trust bar is non-negotiable). This is **HITL**, never auto-graduated.\n- **Fail the recurrence check** → stays ephemeral. Do not graduate on a single run.\n\nGraduation is **never automatic** and is **out of scope for auto-commit** — it is a deliberate,\noperator-approved promotion. A clean simulation that graduated nothing is a normal, complete run.\n\n---\n\n## Human Gate Principle\n\nConvergence within an AI-AI loop is **provisional**. Elevated to final only after human gate.\n\n| Stage | Human gate condition |\n|---|---|\n| Area A S-tier | User reviews + says \"convergence approved\" |\n| Area B convergence | AI-AI loop is provisional → elevated after user confirmation |\n| M-tier auto-fix | Commit after fix (no PR gate); PR requires explicit user request |\n| Three-Doctor Loop end | User reviews convergence and decides next action |\n\n---\n\n## Done When\n\n| Condition | Verdict |\n|---|---|\n| `sim-conductor [Area X] complete` format output | ✅ Simulation complete |\n| 1+ M-tier → fixed + committed (or \"none\") | ✅ Prescription complete |\n| Report `tracks\u002F_meta\u002Fsim_YYYY_MM_DD_*.md` saved | ✅ Persistence complete |\n| 0 M-tier → report committed + S\u002FR backlog reported | ✅ Health check complete |\n| risk≥medium → `cross_model_coverage` recorded in report (external\u002Fcross-session\u002FNONE) | ✅ Coverage gate ran *(check class: measured — the recorded value reflects the dispatch path that ran, not a self-grade; pair: NONE withdraws auto-commit per Step 4)* |\n| Synthesizer emitted the zero-coverage map (every profile standpoint marked covered\u002FZERO) | ✅ Blind-spot surface reported *(check class: mechanical — a checklist over the profile, not a judgment)* |\n\nVerdicts: PASS · CONDITIONAL_PASS (S\u002FR only, or Area B cadence skip) · FAIL (M-tier unresolved) · ESCALATE (persona conflict requiring human judgment, **or** `cross_model_coverage: NONE` on risk≥medium → auto-commit withdrawn pending operator go).\n\n**Mandatory for Area A (external publish)**: steel-quench must complete in same session before Area A is marked complete.\n\n---\n\n## Simplification Guard\n\n- Area B 2+ times in one week: check whether previous S-tier resolved first\n- New simulation without processing M-tier: warn about unresolved M-tier\n- Add new persona: require 1+ empirical cases not captured by current palette\n\n---\n\n## Operations Notes\n\n**AI-AI Loop Bias Defense**: sim-conductor → harness-pr-reviewer → sim-conductor loop shares LLM cognitive blind spots. Internal convergence is \"provisional convergence\" — elevated only by external review or human gate. Area A\u002FB convergence without human gate = incomplete.\n\n---\n\n## Three-Doctor Loop Integration\n\nharness-doctor (structure) · context-doctor (context) · sim-conductor (future behavior) form a closed loop.\n\n| Situation | Next skill |\n|---|---|\n| Structural problem in simulation results | `\u002Fharness-doctor` |\n| Token waste pattern | `\u002Fcontext-doctor` |\n| All three mentioned simultaneously | Three-Doctor Loop: diagnosis→prescription→re-diagnosis cycle |\n| Area A before external publish | `\u002Fsteel-quench` **required** first — mandatory gate |\n\n---\n\n## Path B (External Environment)\n\nWhen `~\u002Fsim\u002F` absent: fallback to harness root, logical isolation only.\n\n> **Detail**: See `SKILL_detail.md §PathB-Detail` — git fallback, logical isolation prompt directive, Area restriction rules — read when running in external\u002Frestricted environment.\n","# sim-conductor — Meta-Simulation Automation Orchestrator\n\n> Profiles target → derives personas → dispatches parallel agents → M\u002FS\u002FR triage → commit.\n> Personas are sourced installed-first → **filled from the persona-container schema** (7 slots as of 2026-07-18 — stance added, not an\n> ad-hoc directive) → external install. Scale: 3 to 16, extensible to a crowd behind the\n> marginal-coverage stop (§Scale).\n\n## Invocation Triggers\n\n| Trigger pattern | Suggested Area |\n|---|---|\n| \"Look at this from an external user's perspective\", \"How would someone else see this?\" | Area A |\n| \"Check the harness\", \"asset audit\", \"internal audit\" | Area B |\n| \"Any new ideas?\", \"ideation scan\", \"naming candidates\" | Area C |\n| \"Review this code\" (pre-PR), \"review this prompt\" | Area D (code) |\n| \"Check if the session card is correct\", \"does the skill actually work\", \"memory cold-start validation\" | Area D (consumer) |\n| \"Is the output good?\", \"artifact quality review\" | Area E |\n\nProposal format: `\"If it's related to [X], should I simulate with \u002Fsim-conductor [Area]?\"`\n\n### Natural Language Triggers\n\n| Example utterance | Intent | Mapped Area |\n|---|---|---|\n| \"Validate our system virtually\" | Full system simulation | Area B |\n| \"Test it with personas\", \"Run an external user simulation\" | External user reaction | Area A |\n| \"Look at it through someone else's eyes\" | External perspective | Area A |\n| \"Validate that this actually works\" | Real-usage validation | Area D |\n| \"Find problems aggressively\" | Adversarial validation | Area B (`challenger`) |\n\n## Triggers\n\n```\n\u002Fsim-conductor                        # default: Area C auto-run\n\u002Fsim-conductor A                      # external user simulation\n\u002Fsim-conductor B                      # internal meta audit\n\u002Fsim-conductor C                      # Innovator scan\n\u002Fsim-conductor all                    # full A + B + C\n\u002Fsim-conductor A --target \u003Cpath>      # specific asset target\n\u002Fsim-conductor D --target \u003Cfile>      # code\u002Fprompt multi-persona review (pre-PR)\n\u002Fsim-conductor D session [--target]   # session card cold-start validation\n\u002Fsim-conductor D skill \u003Cname>         # skill trigger utterance consumer simulation\n\u002Fsim-conductor D memory \u003Cfile>        # memory file cold-start validation\n\u002Fsim-conductor E --target \u003Cfile>      # artifact quality review (post-pipeline)\n```\n\n**Frequency limits**: Area A after asset changes or manual trigger · Area B **min 7-day interval** (executor-applied guard: stop a re-run inside 7 days) · **event-bound, no periodic cadence** (operator decision 2026-10-19): run Area B before a minor\u002Fmajor release or a public announcement — the moment an outside reader's view pays — unless one already ran since the previous minor release. The trigger lives in `templates\u002FPRE-PUBLISH-CHECKLIST.md §Read it as a reader`; `fh_audit_check.zsh` no longer nags on a 30-day clock · Area C\u002FD\u002FE no limit. Why: per-change 4-axis gates with a cross-family reviewer, the nightly live-eval and the weekly harvest cover what a monthly whole-plugin pass used to; what only Area B sees (whole-set naming gaps, an outsider's first read) matters at publication.\n\n---\n\n## Step 0 — Environment Sync\n\nAuto-detect harness root, sync sim clone if present, set `$REPORT_DIR`.\n\n> **Detail**: See `SKILL_detail.md §Step0-Bash` — harness root detection, sim clone sync, report dir creation — read when executing Step 0.\n\n**Area B frequency check** (the 7-day *minimum* interval): Before running Area B, inspect the latest Area B report in `$REPORT_DIR`; if it is less than 7 days old, stop and report its date. This is an executor-applied procedure, not an automatic shell gate. No report in the selected `$REPORT_DIR` → treat as first run for this check. There is no maximum-interval reminder any more — when to run is decided by the release\u002Fannouncement event above.\n\n---\n\n## Step 0.3 — Target Profile Analysis\n\n> **Schema**: `knowledge\u002Fshared\u002Fharness-core\u002Ftpa_schema.md` — canonical field definitions, artifact type table, gate routing. Use §Profile Output Format for output block.\n\nRuns when Area is unspecified, or when target file is provided before Area selection.\n**Skip** if user provides a complete directive (e.g. `\u002Fsim-conductor A --target \u003Cfile>`).\n\nRead target artifact(s) → classify on 5 dimensions → output recommendation → user confirms or overrides:\n\n| Dimension | Signal → Weight shift |\n|---|---|\n| `artifact_type` | SKILL.md \u002F design-doc → Area B + D-skill↑ · README \u002F CHEATSHEET → Area A↑ · code \u002F config → Area D-code↑ |\n| `audience` | external installer \u002F first-time user → beginner↑ · internal team only → challenger↑ |\n| `claim_density` | 3+ stated benefits or superlatives → challenger↑ |\n| `risk_level` | external publish \u002F marketplace listing → steel-quench prerequisite triggered. **Mechanical floor (not judge-only)**: any of — publish\u002Fmarketplace target · public-surface or visibility change · auth\u002Fsecret-handling or executable code · an FH asset under the 4-axis gate — **forces `risk_level ≥ medium`** regardless of profiler judgment. The floor closes the \"fool the profiler into `low` to skip Step 0.6\" seam; the judge may only raise above the floor, never below it. |\n| `novelty` | first-of-its-kind \u002F no prior session evidence → phantom-quench recommended |\n\n```\nTarget Profile output:\n  artifact_type: [type]\n  audience: [internal | external | mixed]\n  claim_density: [low | medium | high] ([N] stated claims)\n  risk_level: [low | medium | high]\n\nRecommendation:\n  Areas: [list + rationale]\n  Persona composition: [list + weight]\n  Scale: [Minimum 3 | Extended 4–8 | Full ≤16]\n  Prerequisites: [steel-quench \u002F phantom-quench \u002F none]\n```\n\n#### Persona Discovery (after profile → before dispatch)\n\n> **Schema**: `knowledge\u002Fshared\u002Fharness-core\u002Fpersona_container_schema.md` — the 7-slot persona-definition\n> container. ② below **fills** these slots (lens \u002F internal-logic \u002F external-grounding \u002F output-protocol \u002F stance\n> \u002F cost-tier \u002F lifecycle) instead of injecting a shapeless directive. The container is what makes a\n> synthesized persona reproducible and cost-routable; the old \"fallback palette\" was undefined.\n\nAfter profiling, sim-conductor determines the needed perspective types, then maps each to the best available agent:\n\n```\nFor each needed perspective:\n  ① Scan installed plugins + .claude\u002Fagents\u002F → exact match? → use it\n  ② Fill the 7-slot persona container — lens · internal-logic · external-grounding · output-protocol · stance\n       (= Step 1.5 parallax shape) · cost-tier · lifecycle → a defined, reproducible persona; the filled\n       container IS the dispatch prompt (replaces the old ad-hoc \"fallback palette\" directive). Assign\n       the cost-tier slot per §Scale cost-tier routing. [Schema: persona_container_schema.md]\n  ③ GAP: no ①② match for a high-weight perspective\n       → query \u002Fplugin-recommender: \"find agents for [perspective] matching [artifact_type] context\"\n       → present: name · install command · estimated fit · token cost\n       → user chooses: install now \u002F skip \u002F substitute with ②\n```\n\nPersona Discovery output:\n\n```\nPersona Map:\n  beginner       → [installed agent: beginner] OR [filled 7-slot container]\n  challenger     → [installed agent: challenger] OR [filled 7-slot container]\n  [profile-specific role] → ⚠️ GAP — plugin-recommender recommends: [X] (install? y\u002Fn)\n```\n\nGaps on **high-weight personas** (from profile) block dispatch until resolved. Gaps on low-weight personas → auto-fill with built-in fallback, no block.\n\n**Degraded coverage rule**: When user skips a high-weight gap (substitutes with built-in fallback), flag in the Step 3 report under \"Persona composition used\" as `⚠️ degraded: [perspective]`. Do not silently proceed. Degraded coverage on risk_level=high targets → additionally warn before dispatch.\n\nOverriding a prerequisite recommendation requires explicit \"skip [prerequisite]\" utterance.\n\n> **Detail**: See `SKILL_detail.md §PersonaDiscovery` — plugin-recommender query format, gap resolution decision tree, persona map examples for each artifact type — read when running Persona Discovery.\n\n> **Detail**: See `SKILL_detail.md §Profile-Examples` — worked profiles for SKILL.md, README, bash script, design doc — read when result is ambiguous or to calibrate weights.\n\n---\n\n## Step 0.5 — Precondition Validation\n\nProactively surface 1 concern, then confirm whether to proceed. Skip if no concern.\n\n| Check | Concern trigger condition |\n|---|---|\n| Area B unresolved S-tier | An unresolved S-tier finding from the previous run |\n| Unresolved M-tier | New simulation without processing previous M-tier |\n| steel-quench not run | Area A in external publish context → **mandatory gate: Area A does not proceed until steel-quench completes** |\n| Self-recursive path | sim-conductor attempting to auto-PR its own M-tier |\n| Goal-method alignment | Requested Area doesn't match actual problem |\n| D\u002FE --target missing | Area D(code)\u002FE runs without --target file |\n| D session target not found | Session start card file not found → confirm path |\n| E execution timing error | Area E before pipeline run (no artifact yet) |\n\nConcern format: `\"One thing to check before [Area X]: [concern]. Proceed?\"`\n\n---\n\n## Step 0.6 — Cross-Model Coverage Gate (risk≥medium — hard)\n\nCloses the homogeneous-blind-spot + formatting-flattery vector (judge-robustness swarm, 2026-06-13):\na panel of same-session Claude sub-agents shares one model's blind spots, so a clean verdict can be\nflattery the **whole panel** is blind to — and with no quotable-rule violation, no persona escalates.\nFor `risk_level ≥ medium` targets (from Step 0.3), at least one persona MUST come from a source\n**outside the orchestrator's own session context**. This **promotes the former advisory \"dual\nvalidation principle\"** (detail §AreaB-Baseline #4) to a hard gate — the mechanical-anchor pattern of\nhardening #1–#5: a judged verdict binds to a fact the judge cannot fake.\n\n**Graceful-degradation ladder** — take the highest available rung. The gate **never breaks\nsim-conductor**; at the bottom it only withdraws the *unsafe autonomy* (self-certifying a blind verdict),\nnot the run:\n\n| Rung | Source | `cross_model_coverage` | Closes | When |\n|---|---|:---:|---|---|\n| 1 | External CLI team (Multi-Team Mode — §MultiTeam) | `external` | model-level blind spot (genuine cross-model) | 1+ external CLI live + probe non-empty |\n| 2 | Cross-session Claude — `claude -p` headless, or an Agent with **zero inherited context** | `cross-session` | **session-contamination only** — a fresh Claude shares the same weights\u002FRLHF gradient, so it does **not** close the model-level blind spot; it only removes the orchestrator's working-memory bias. Honest partial mitigation, labeled as such | no external CLI; dispatch probe returns non-empty |\n| 3 | Same-session sub-agents only | `NONE` | nothing — homogeneous panel | neither rung's probe succeeded |\n\n**Mechanical anchor** — `cross_model_coverage` is valid **only if backed by a quoted dispatch artifact**,\nnot a self-assessment (the self-signing hole hardening #1 closed for the marker — same fix here). To\nrecord `external` or `cross-session`, the Step 3 report must **quote a non-empty excerpt of the actual\ndispatch output** (external CLI stdout, or the dispatched Agent's returned verdict text); a label with no\nquoted excerpt is invalid and falls to `NONE`. **Liveness, not mere availability**: probe the rung before\nclaiming it — attempt the dispatch with a timeout; if it errors or returns empty (plan-gate closed,\ncontext saturated, CLI present-but-dead), record `NONE`, never assume the rung succeeded. This is the\nsame honest scope as #1: the artifact makes the claim **auditable**, not cryptographically unforgeable —\na fabricated excerpt remains the operator's + weekly-audit's residual by design. On rung 3 (`NONE`) for a\nrisk≥medium target:\n- the report flags `⚠️ cross-model coverage: NONE — homogeneous same-session panel; verdict provisional`, **and**\n- **Step 4 auto-commit privilege is withdrawn** (see Step 4): M-tier fixes may be *prepared* but the\n  commit waits for the operator's explicit go. Auto-committing a structurally self-blind verdict is\n  exactly what the exploit targeted — so that single privilege is what degrades, not the simulation.\n\n`risk_level = low` targets are exempt (a homogeneous panel is acceptable); the gate fires only at\nmedium+. The rung-2 fallback is what makes this CC-only-safe: a Claude-only environment still gets a\nreal cross-context read (a fresh isolated dispatch shares no working memory with the orchestrator), so\n`NONE` is reached only when *both* external CLIs and a second Claude context are unavailable.\n\n---\n\n## Step 1 — Area-Specific Simulation\n\n### Area A — External User Perspective\n\nTarget: skill descriptions, README, CHEATSHEET entry points.\n\n#### Task-Adaptive Persona Selection\n\nsim-conductor does **not** run a fixed persona set. It derives needed perspectives from the task, then sources each persona in priority order:\n\n```\n① Installed first — scan installed plugins + .claude\u002Fagents\u002F for a matching persona\u002Fagent\n② Container fill — fill the 7-slot persona container for the standpoint; the filled container IS the\n     dispatch prompt into a general-purpose Agent (Path B — replaces the old ad-hoc role directive)\n③ External fetch — chain to \u002Fplugin-recommender when ①② insufficient for high-stakes tasks\n```\n\nShipped standpoint agents (① tier — sourced installed-first):\n\nFH ships a coherent **user-mastery spectrum** as real, reusable agents (not prompt-directive shells) — found by the ① installed-first scan, reusable across skills, and each isolated-context dispatchable for a true cold read (the bias-isolation value: an evaluator outside the author's context reads cold):\n\n| Agent | Spectrum tier | Standpoint | Type |\n|---|---|---|---|\n| `beginner` | entry | First-contact cold-read — onboarding friction a fluent author cannot feel | reasoning |\n| `main-player` | core | Engaged user; intelligently scopes Light \u002F Midcore \u002F Heavy (Heavy = classic power-user edge\u002Flimit lens) | reasoning |\n| `expert` | frontier | Domain authority; web-grounded accuracy + SOTA currency, citation-enforced | data (WebSearch\u002FWebFetch) |\n| `challenger` | adversarial axis | Frontier adversary; U1 absorbs the skeptic \"why not just X?\" lens | adversarial |\n\n> **Lineage**: `beginner` \u002F `main-player` \u002F `expert` are the FH-native frontier successors to the field deep-insight `user` group (newcomer \u002F power-user) — re-derived to FH grade with embedded methodology + Done-When, not name-copied. `challenger` is the advanced form of the field `devil-advocate`. The former standalone skeptic standpoint is folded into `challenger` U1.\n\n**Ad-hoc roles** (② tier — container fill): when the profile demands a standpoint with no shipped agent (e.g. \"security auditor\", \"non-native reader\"), fill the 7-slot persona container for that standpoint — the filled container is the dispatch prompt. Prefer ① shipped agents; use ② only for genuinely task-specific one-offs.\n\n#### Scale\n\n| Scale | Count | When |\n|---|---|---|\n| **Minimum** | 3 | Routine Area A — most task-relevant 3 perspectives |\n| **Extended** | 4–8 | High-stakes publish \u002F external release |\n| **Full** | Up to 16 parallel | Pre-major-version \u002F architecture review |\n| **Crowd** | 16+ — **no fixed cap; bounded by the marginal-coverage stop** | Only when the marginal-coverage stop has demonstrably **not** fired at 16 — i.e. persona #17 is still decorrelated and finding-productive. **Not** chosen by artifact size or audience count (Full already covers those); the stop, not a headcount, opens this tier |\n\n**Crowd is output-driven, never a target** (per the container schema): the 16-cap lifts only behind the\n**marginal-coverage stop** — add a persona only if it is *decorrelated* (a distinct viewpoint, or — at\ncrowd scale — a different model family) from those already running; **stop when the marginal new-finding\nrate → 0**. A crowd of identical-output personas is decorative (steel-quench Wave-1 angle #1). Do not\nname a headcount goal; the count is whatever the stop allows.\n\n**Cost-tier routing (the economy lever)** — each persona's `cost-tier` slot routes it to a model so\nbreadth is cheap and the verdict stays trustworthy:\n- **floor-local \u002F heavy-local** (e.g. a local Ollama sidecar) carry cheap *breadth* — extra decorrelated\n  standpoints at near-zero token cost. A heavier local (27–32B) is a stronger canary than a floor model,\n  but **on one probe, not a benchmark** (`[[persona-container-schema]]` §tier-distribution) — so a local\n  tier is a **canary, never the gate**.\n- **frontier** carries the **terminal verdict** — the M\u002FS\u002FR triage and any auto-commit decision stay\n  frontier. This is the same anchor as Step 0.6: local breadth *decorrelates*, it does not *decide*.\n- Cost-tier routing and the Step 0.6 cross-model gate are complementary, not duplicate: 0.6 *requires* a\n  non-session source for risk≥medium (blind-spot closure); cost-tier *distributes* breadth for economy.\n  A local-sidecar persona can satisfy both at once (cross-family **and** cheap), but only Step 0.6's\n  quoted-artifact rule lets it count as `external` coverage.\n\nAll personas run as **parallel Agents** — no sequential bottleneck. Use agent-composer Medium\u002FLarge\nfan-out for Extended\u002FFull; Crowd scale dispatches in decorrelated waves, re-checking the marginal-coverage\nstop between waves rather than firing all at once.\n\n**Simplification guard**: Routine internal audits → ② built-in fallback at Minimum scale. Chain to ③ only when a needed perspective has no ①② match, or stakes are high.\n\n#### Multi-Team Mode (when external CLIs available)\n\nWhen 1+ external CLIs detected, Area A upgrades to Multi-Team Panel: each CLI forms an independent team. Cross-team: 2+ teams flag same issue → tier-up (S→S-confirmed). External-only findings → \"Claude blind spot\" flag.\n\nPre-entry user confirmation required before multi-team execution.\n\n> **Detail**: See `SKILL_detail.md §MultiTeam` — team formation table (T0–T4), CLI detection bash, confirmation dialog, cross-team synthesis format — read when multi-team mode activates.\n\n**A-1** (`beginner`) — first-contact friendliness · onboarding friction · terminology clarity\n**A-2** (`main-player`) — engaged-use fit; Heavy tier: install conflicts · duplication · silent overwrite\n**A-3** (`challenger`, artifact_type=\"SKILL\") — claim-evidence gaps · angles U3, U5, S2\n\n> ⚠️ **Human review gate**: Area A S-tier judgments require owner review before entering AI-AI loop.\n\n---\n\n### Area B — Internal Meta Audit\n\nTarget: all fh-meta assets (skills + agents + plugin.json).\n\n> **Self-recursive caution**: When M-tier found in sim-conductor itself → report only, delegate to user.\n\n**Dispatch pattern**: hub-persona-auditor ∥ persona-innovator (parallel) → challenger (after, with both outputs as context)\n\n1. **hub-persona-auditor** — README\u002FCHEATSHEET as external briefing, 4-axis review, 3-tier suggestions\n2. **persona-innovator** (Mode I) — naming gaps + structural gaps + 3-5 candidates\n3. **challenger** (artifact_type=\"SKILL\") — receives 1+2 outputs, angles U2, D3, U5 — \"what was missed?\"\n\nRationale: 1+2 run without influencing each other; challenger synthesizes both perspectives independently, avoiding Cost of Consensus.\n\n> **Detail**: See `SKILL_detail.md §AreaB-Baseline` — external baseline injection methods, dual-validation principle, Area B → steel-quench handoff — read when setting up Area B defenses.\n\n> ⚠️ **Human review gate**: Area B convergence is provisional until owner confirms.\n\n---\n\n### Area C — Innovator Scan\n\nRuns independently. Skip if included in Area B.\n\n**persona-innovator** (Mode F = Internal + External): current harness gaps + frontier scan. Output: naming candidates + absorption signals + top 1 priority action.\n\n---\n\n### Area D — Artifact Validity Validation\n\n| Mode | Trigger | Core question |\n|---|---|---|\n| **code** | `\u002Fsim-conductor D --target \u003Cfile>` | Is the code\u002Fprompt well-built? |\n| **consumer** | `\u002Fsim-conductor D session\u002Fskill\u002Fmemory` | Does the artifact work for the intended consumer? |\n\n#### D-code — Profile-Aware Persona Routing\n\nPersona composition adapts to `artifact_type` from Step 0.3 profile:\n\n| Artifact type | Primary persona | Supporting persona | Focus |\n|---|---|---|---|\n| SKILL.md \u002F design doc | challenger (artifact_type=\"SKILL\") | `beginner` | Governance gaps, behavioral rule coverage |\n| Python \u002F JS \u002F bash code | challenger (artifact_type=\"Code\") | `main-player` (Heavy) | Edge cases, performance, security surface |\n| Prompt \u002F config | `beginner` | challenger | Interpretation errors, implicit assumptions |\n| Auth \u002F security-sensitive | challenger + Security-auditor† | `main-player` (Heavy) | Attack surface, privilege escalation |\n\n† Security-auditor = ② tier — the 7-slot persona container filled for a security-auditor standpoint.\n\nAll personas run in parallel. Findings = M\u002FS\u002FR → M-tier items fixed immediately.\n\n#### D-consumer — Artifact Consumer Simulation\n\nConsumer agent attempts actual use (not just reads and judges). Grades: F (functional) \u002F P (partial → S-tier) \u002F B (broken → M-tier immediate fix).\n\n> **Detail**: See `SKILL_detail.md §AreaD-Consumer` — D-session 4 questions, D-skill cold-start procedure, D-memory check items, verdict criteria — read when executing D-consumer.\n\n---\n\n### Area E — Artifact Quality Review\n\n`expert` objection (E-1) + Practitioner confusion (E-2) in parallel → Pattern structuring (E-3) integrates both.\n\n> **Detail**: See `SKILL_detail.md §AreaE-Detail` — E-1\u002FE-2\u002FE-3 execution, finding format, pattern naming procedure — read when executing Area E.\n\n---\n\n## Step 1.5 — Persona Output Protocol + Neutral Synthesizer (parallax)\n\nGeneralized from the field `deep-insight` multi-persona pattern (private companion store), domain-stripped — the *pattern*\nis renamed **parallax** for public FH (it is a mode of this skill, not a separate skill — see asset-placement\n2026-06-06). It gives the persona dispatch above a shared output contract + a neutral aggregator, so\nmulti-persona findings stay comparable and the synthesis injects no bias of its own.\n\n> **Naming provenance (precise)**: \"renamed\" above refers to the *pattern* (→ parallax), not the personas.\n> The company-team-coupled field personas (fe\u002Fbe\u002Fios\u002Fpm\u002Fux-writer\u002Fcompliance\u002Fqa) were domain-stripped\n> entirely. The generic `user` group (newcomer\u002Fpower-user) was **re-derived to FH grade as the shipped\n> `beginner` \u002F `main-player` \u002F `expert` mastery-spectrum agents** (embedded methodology + Done-When, not\n> name-copied shells), and the field `devil-advocate` was **advanced into `challenger`** (sandboxed-adversary\n> + adaptive attack matrix). Lineage is acknowledged; nothing is carried verbatim as a shell.\n\n**Shared persona output protocol** — every dispatched persona emits the same shape, whatever its lens:\n\n```\n### Strengths        (0–3, from this persona's viewpoint)\n### Concerns\n  Critical   — compile\u002Fruntime failure · clear logic error · data corruption · security leak\n  Important  — significant user\u002Fservice impact in a plausible scenario\n  Suggestion — optional improvement\n  (each item: [file:line or quoted span] one-line summary — rationale)\n### Open questions   (0–3 items needed for a decision)\n### Absence check    (outside-vantage personas — beginner\u002Fintegrator: what does the artifact FAIL to\n                      specify that this standpoint needs? discoverability · undocumented contract ·\n                      unstated assumption. A normal, self-administrable rubric item — surfaces real gaps.)\n```\n\n**FP judgment discipline** — only escalate when confident. Never escalate: pre-existing issues not\nintroduced by this change · style\u002Fquality without a quotable rule · linter-catchable · speculative\n(\"might break if…\") · subjective preference (→ Suggestion). **If not confident, do not mark it** — false\npositives erode reviewer trust.\n\n**Neutral synthesizer** — the aggregator is a NON-persona; it adds no opinion of its own:\n- No opinion injection — never a conclusion no persona stated.\n- Preserve attribution — always traceable which persona said what.\n- Priority labels verbatim — Critical\u002FImportant\u002FSuggestion carried as the persona set them.\n- No forced consensus or forced conflict — report Common opinions (2+ personas agree) and Conflicts\n  (position A vs B, each with rationale) as-is. Feeds Step 2 M\u002FS\u002FR triage (M ← Critical or 2+ personas).\n- Cost-tier aware triage — a frontier-tier persona's verdict is terminal; a floor\u002Fheavy-local persona's\n  finding enters only as decorrelated **breadth** (it can raise a standpoint to `covered`, never solely\n  set an M-tier on its own). This enforces in the aggregator the §Scale \"local breadth decorrelates, it\n  does not decide\" lock — without it, a local canary's finding would drive triage identically to frontier.\n\n**Zero-coverage map (mandatory synthesizer output)** — the synthesizer must emit, mechanically, the set\nof standpoints that produced **no** finding, not only the ones that did (judge-robustness swarm,\n2026-06-13). Enumerate the persona **standpoints** in play — those the Step 0.3 profile recommended,\n**plus the standpoints its dimensions imply** (risk_level=high → a security\u002Fpublish standpoint ·\naudience=mixed → a non-native-reader standpoint · claim_density=high → a claim-evidence standpoint).\nList the *standpoints*, not the bare dimension names (a `risk_level (low) → ZERO` row is noise that\ntrains operators to ignore ⚠️). Mark each `covered` (≥1 persona addressed it) or `ZERO` (no persona\ntouched it):\n\n```\nCoverage map:\n  beginner (onboarding friction)      → covered (A-1: 2 findings)\n  challenger (claim-evidence)         → covered (A-3: 1 finding)\n  security surface (risk_level=high)  → ZERO ⚠️\n  non-native reader (audience=mixed)  → ZERO ⚠️\n```\n\nA clean report with `ZERO` standpoints is **not** a pass — it is an uncovered surface, reported as such.\nThis converts the formatting-flattery failure (everything reads fine → nothing escalates) into a visible\ngap: **silence on a standpoint is reported as `ZERO`, never inferred as approval.** It is a checklist\nderived from the profile and the dispatch outputs — a mechanical anchor, not a judgment. Carry the map\nverbatim into the Step 3 report.\n\nThe two severity vocabularies are layered, not redundant: a persona running **in isolation** assigns only\nits own Critical\u002FImportant\u002FSuggestion — it cannot assign M\u002FS\u002FR, since `S = found by 3+ personas` depends on\ncross-persona agreement the isolated persona never sees. The synthesizer is the only context that can triage\nto M\u002FS\u002FR. So isolation *requires* the per-persona → synthesized two-layer split.\n\n**External-harness persona sourcing** — isomorphic to steel-quench Step 0.4 (Specialized Reviewer\nDiscovery) + Wave 5 (external CLI teams). A needed lens may be sourced from an **installed sibling\nharness**, not only the built-in palette — e.g. gstack `\u002Freview` (staff-engineer), `\u002Fcso`\n(security-officer), `\u002Fqa` (QA-lead) when gstack is installed. sim-conductor orchestrates; the sibling\nsupplies the specialist lens. Same ①installed → ②fallback → ③fetch priority as Persona Discovery — an\nexternal harness's review-skills count as ① installed sources, widening the persona pool without FH\nshipping every specialist.\n\n> **Absence check — resolved (added above)**: the clean replication (companion-store experiment, Arm F: identical\n> artifact, explicit omission prompt, ownership-only variable) found self ≈ isolated (~90% overlap) —\n> **omission-detection is self-administrable when explicitly asked**, refuting the earlier \"the author\n> can't see their own omissions\" (Arm E, a prompt\u002Fdesign-drift artifact). So the Absence check is a\n> normal, valuable rubric item (it surfaces real undocumented-contract gaps), not a special isolation-only\n> power. The pattern's value is rubric\u002Fstandpoint *supply + routine enforcement* (copyable utility), not\n> de-biasing — see `knowledge\u002Fshared\u002Fpatterns\u002Fmulti-persona-review.md`.\n\n---\n\n## Step 2 — Synthesis\n\n| Tier | Criteria | Action |\n|---|---|---|\n| **M (Mandatory)** | Blocks external user entry \u002F naming-substance mismatch \u002F security\u002Ftrust damage | Immediate fix required |\n| **S (Strong)** | Feature degradation \u002F found by 3+ personas | Address within next session |\n| **R (Recommended)** | Improvement value \u002F found by single persona | Backlog |\n\n**Deduplication**: Multiple personas → same location: M > S > R.\n\n**Early Convergence Detection**: 2+ Areas independently classify same item as M-tier → immediate fix, remaining Areas continue for S\u002FR coverage only.\n\n---\n\n## Step 3 — Report\n\nFile: `$REPORT_DIR\u002Fsim_YYYY_MM_DD_area_[X].md`\n\n> **Detail**: See `SKILL_detail.md §Report-Format` — full report template with frontmatter, M\u002FS\u002FR sections, emerging asset candidates — read when writing the report.\n\n---\n\n## Step 4 — PR or Direct Commit\n\n1+ M-tier → fix immediately → commit. PR creation requires explicit user request.\n0 M-tier → commit report only + report S\u002FR backlog.\n\n**Cross-model gate on auto-commit (risk≥medium)** — when Step 0.6 recorded `cross_model_coverage: NONE`\non a risk≥medium target, the auto-commit privilege is **withdrawn**: prepare the M-tier fixes and write\nthe report, but do **not** self-commit — surface *\"cross-model coverage NONE on a risk≥medium target;\nthe verdict is from a homogeneous same-session panel. Commit the fixes, or add a cross-context read\nfirst?\"* and wait for the operator's go. `external`\u002F`cross-session` coverage, or risk_level=low, commits\nas normal. (This withdraws one privilege, not the run — the report and fixes still exist.)\n\n> **Detail**: See `SKILL_detail.md §PR-Bash` — branch creation bash, commit + push, gh pr create template — read when creating a PR.\n\n---\n\n## Step 5 — Graduate a Proven Persona (optional, rare)\n\nMost synthesized personas are **throwaway** — the run accelerated the work and the persona is discarded.\nA persona earns a permanent home only by passing the container schema's **admission test** (the analogue\nof \"app-ification\"; `[[persona-container-schema]]` §Lifecycle):\n\n| Check | Criterion | Class |\n|---|---|---|\n| **Recurrence** | surfaced **≥1 unique S\u002FM finding** (one no other persona in the run produced) across **≥2 distinct targets** — not a one-off lucky fire | measured |\n| **Trust bar (skills only)** | if graduating into a *skill* rather than a project's persona set, additionally clears the **FH 6-item creation gate** | mandatory-pass |\n\n- **Pass → into a project's static set**: route the write through `asset-placement-gate` (expected\n  verdict: project-local) before writing the filled container to the project's persona file (e.g.\n  the-bible `core\u002Fpersonas.json`) — sim-conductor proposes the target, the placement gate confirms it,\n  the operator approves. A project-local asset, not an FH change, but the cross-boundary write is still gated.\n- **Pass → into a skill**: route through `asset-placement-gate` + the New-Skill creation gate (the\n  skill-foundry inversion — the trust bar is non-negotiable). This is **HITL**, never auto-graduated.\n- **Fail the recurrence check** → stays ephemeral. Do not graduate on a single run.\n\nGraduation is **never automatic** and is **out of scope for auto-commit** — it is a deliberate,\noperator-approved promotion. A clean simulation that graduated nothing is a normal, complete run.\n\n---\n\n## Human Gate Principle\n\nConvergence within an AI-AI loop is **provisional**. Elevated to final only after human gate.\n\n| Stage | Human gate condition |\n|---|---|\n| Area A S-tier | User reviews + says \"convergence approved\" |\n| Area B convergence | AI-AI loop is provisional → elevated after user confirmation |\n| M-tier auto-fix | Commit after fix (no PR gate); PR requires explicit user request |\n| Three-Doctor Loop end | User reviews convergence and decides next action |\n\n---\n\n## Done When\n\n| Condition | Verdict |\n|---|---|\n| `sim-conductor [Area X] complete` format output | ✅ Simulation complete |\n| 1+ M-tier → fixed + committed (or \"none\") | ✅ Prescription complete |\n| Report `tracks\u002F_meta\u002Fsim_YYYY_MM_DD_*.md` saved | ✅ Persistence complete |\n| 0 M-tier → report committed + S\u002FR backlog reported | ✅ Health check complete |\n| risk≥medium → `cross_model_coverage` recorded in report (external\u002Fcross-session\u002FNONE) | ✅ Coverage gate ran *(check class: measured — the recorded value reflects the dispatch path that ran, not a self-grade; pair: NONE withdraws auto-commit per Step 4)* |\n| Synthesizer emitted the zero-coverage map (every profile standpoint marked covered\u002FZERO) | ✅ Blind-spot surface reported *(check class: mechanical — a checklist over the profile, not a judgment)* |\n\nVerdicts: PASS · CONDITIONAL_PASS (S\u002FR only, or Area B cadence skip) · FAIL (M-tier unresolved) · ESCALATE (persona conflict requiring human judgment, **or** `cross_model_coverage: NONE` on risk≥medium → auto-commit withdrawn pending operator go).\n\n**Mandatory for Area A (external publish)**: steel-quench must complete in same session before Area A is marked complete.\n\n---\n\n## Simplification Guard\n\n- Area B 2+ times in one week: check whether previous S-tier resolved first\n- New simulation without processing M-tier: warn about unresolved M-tier\n- Add new persona: require 1+ empirical cases not captured by current palette\n\n---\n\n## Operations Notes\n\n**AI-AI Loop Bias Defense**: sim-conductor → harness-pr-reviewer → sim-conductor loop shares LLM cognitive blind spots. Internal convergence is \"provisional convergence\" — elevated only by external review or human gate. Area A\u002FB convergence without human gate = incomplete.\n\n---\n\n## Three-Doctor Loop Integration\n\nharness-doctor (structure) · context-doctor (context) · sim-conductor (future behavior) form a closed loop.\n\n| Situation | Next skill |\n|---|---|\n| Structural problem in simulation results | `\u002Fharness-doctor` |\n| Token waste pattern | `\u002Fcontext-doctor` |\n| All three mentioned simultaneously | Three-Doctor Loop: diagnosis→prescription→re-diagnosis cycle |\n| Area A before external publish | `\u002Fsteel-quench` **required** first — mandatory gate |\n\n---\n\n## Path B (External Environment)\n\nWhen `~\u002Fsim\u002F` absent: fallback to harness root, logical isolation only.\n\n> **Detail**: See `SKILL_detail.md §PathB-Detail` — git fallback, logical isolation prompt directive, Area restriction rules — read when running in external\u002Frestricted environment.\n",{"name":5,"model-note":23,"description":7,"allowed-tools":24,"user-invocable":31},"session-inherit — Sonnet base is first-class (sonnet_floor_doctrine.md); depth-critical judged steps route to dispatch (opus agent \u002F cross-family sidecar, consent-gated), never a substrate requirement",[25,26,27,28,29,30],"Read","Write","Bash","Grep","Glob","Agent",true,"main","899495d17a23f30ca149d4ac29a137a8fa7864e3","plugins\u002Ffh-meta\u002Fskills\u002Fsim-conductor",1,"https:\u002F\u002Fgithub.com\u002Fchrono-meta\u002Fforge-harness","https:\u002F\u002Fchrono-meta.github.io\u002Fforge-harness\u002F","https:\u002F\u002Fgithub.com\u002Fchrono-meta","2026-11-15 02:42:21",{"source":8,"skillName":5,"commands":41,"prompt":51,"zipAvailable":31},[42,45,48],{"tool":43,"command":44},"npx","npx skills add chrono-meta\u002Fforge-harness --skill sim-conductor",{"tool":46,"command":47},"bunx","bunx skills add chrono-meta\u002Fforge-harness --skill sim-conductor",{"tool":49,"command":50},"pnpm","pnpm dlx skills add chrono-meta\u002Fforge-harness --skill sim-conductor","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fchrono-meta\u002Fforge-harness\u002Fblob\u002Fmain\u002Fplugins\u002Ffh-meta\u002Fskills\u002Fsim-conductor\u002FSKILL.md",[53],{"path":54,"resourceType":55,"resourceTypeDesc":56,"sizeBytes":57,"sourceHtmlUrl":58},"SKILL_detail.md","REFERENCE","参考文档",17350,"https:\u002F\u002Fgithub.com\u002Fchrono-meta\u002Fforge-harness\u002Fblob\u002Fmain\u002Fplugins\u002Ffh-meta\u002Fskills\u002Fsim-conductor\u002FSKILL_detail.md",[60,67,74,81,87,94],{"skillNo":61,"name":62,"nameZh":6,"description":63,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":64,"categories":65,"keywords":66,"publishedAt":18},"01M74ZPW36H1HVE0R61QZ5ZND0","todo-orchestrator","TODO CLI 프로젝트의 기능 추가·버그 수정·리팩터링·테스트 보강을 개발자와 QA 에이전트로 조율한다. todo 앱 관련 코드 작업 요청에 사용한다. 다시 실행, 재실행, 수정, 보완, 이전 결과 개선, QA만 다시 같은 후속 요청에도 사용한다. 단순 질문에는 사용하지 않는다.","https:\u002F\u002Fgithub.com\u002Fchrono-meta\u002Fforge-harness\u002Fblob\u002Fmain\u002Fscripts\u002Ffixtures\u002Fgenerated_harness\u002FP4_r2\u002F.claude\u002Fskills\u002Ftodo-orchestrator\u002FSKILL.md",[],[],{"skillNo":68,"name":69,"nameZh":6,"description":70,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":71,"categories":72,"keywords":73,"publishedAt":18},"01M74ZPS4EP2349B346R5JQEYE","todo-dev","TODO CLI(todo\u002Fcli.py)에 명령을 추가하거나 버그를 고치고 테스트를 작성하는 방법. todo 앱 코드 수정·기능 추가·테스트 작성·리팩터링 작업에 사용한다. 수정, 보완, 다시 구현 요청에도 사용한다.","https:\u002F\u002Fgithub.com\u002Fchrono-meta\u002Fforge-harness\u002Fblob\u002Fmain\u002Fscripts\u002Ffixtures\u002Fgenerated_harness\u002FP4_r2\u002F.claude\u002Fskills\u002Ftodo-dev\u002FSKILL.md",[],[],{"skillNo":75,"name":76,"nameZh":6,"description":77,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":78,"categories":79,"keywords":80,"publishedAt":18},"01M74ZPPX2743J6FPT8N2ZT0D5","todo-dev-team","TODO CLI(todo\u002Fcli.py) 개발 작업을 개발자·테스터·리뷰어 팀으로 진행하는 오케스트레이터. 명령 추가, 버그 수정, 테스트 작성, 리팩터링, 커밋·푸시 요청에 사용한다. 다시 실행, 재실행, 수정, 보완, 리뷰 반영, 테스트만 다시 같은 후속 요청에도 사용한다. 단순 질문에는 쓰지 않는다.","https:\u002F\u002Fgithub.com\u002Fchrono-meta\u002Fforge-harness\u002Fblob\u002Fmain\u002Fscripts\u002Ffixtures\u002Fgenerated_harness\u002FP3_r2\u002F.claude\u002Fskills\u002Ftodo-dev-team\u002FSKILL.md",[],[],{"skillNo":82,"name":69,"nameZh":6,"description":83,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":84,"categories":85,"keywords":86,"publishedAt":18},"01M74ZPKWST7DF24H0EN9RCEME","TODO CLI 개발 작업(명령·기능 추가, 버그 수정, 테스트 작성, 리뷰, 커밋·push)을 개발자·테스터·리뷰어 에이전트 팀으로 진행하는 오케스트레이터. todo CLI, todo.json, todo\u002Fcli.py 관련 구현·수정·테스트 요청에 사용한다. 재실행, 수정, 보완, 이전 결과 개선, 테스트만 다시, 리뷰만 다시 요청에도 사용한다. 단순 질문에는 쓰지 않는다.","https:\u002F\u002Fgithub.com\u002Fchrono-meta\u002Fforge-harness\u002Fblob\u002Fmain\u002Fscripts\u002Ffixtures\u002Fgenerated_harness\u002FP3_r1\u002F.claude\u002Fskills\u002Ftodo-dev\u002FSKILL.md",[],[],{"skillNo":88,"name":89,"nameZh":6,"description":90,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":91,"categories":92,"keywords":93,"publishedAt":18},"01M74ZPHBP62X3KKSV2RVKSBKT","weekly-summary-write","정리된 회의록으로 주간 회의 요약 문서를 작성하는 형식과 문체 규칙. 주간 요약 작성, 요약 문서 수정·보완, 검토 의견 반영 시 사용한다.","https:\u002F\u002Fgithub.com\u002Fchrono-meta\u002Fforge-harness\u002Fblob\u002Fmain\u002Fscripts\u002Ffixtures\u002Fgenerated_harness\u002FP2_r2\u002F.claude\u002Fskills\u002Fweekly-summary-write\u002FSKILL.md",[],[],{"skillNo":95,"name":96,"nameZh":6,"description":97,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":98,"categories":99,"keywords":100,"publishedAt":18},"01M74ZPECSXY4C0QNHAD96Z88H","weekly-meeting-summary","주간 회의록을 정리하고 요약 문서를 만드는 전체 작업을 조율하는 오케스트레이터. 회의록 요약, 주간 회의 정리, 주간 보고용 요약 요청 시 사용한다. 다시 실행, 재실행, 요약 수정·보완, 이전 결과를 바탕으로 한 개선, 정리만\u002F검토만 다시 같은 후속 요청에도 사용한다.","https:\u002F\u002Fgithub.com\u002Fchrono-meta\u002Fforge-harness\u002Fblob\u002Fmain\u002Fscripts\u002Ffixtures\u002Fgenerated_harness\u002FP2_r2\u002F.claude\u002Fskills\u002Fweekly-meeting-summary\u002FSKILL.md",[],[]]