refocus
Audits an existing Clean Room run and steers it back to missed gates without expanding declared scope.
- License 许可
- Apache-2.0
- 最近更新
- 2026-11-14
- Stars
- ★ 1.6k
- Forks
- 491
01 前言 Front Matter
- name
- refocus
- description
- Audits an existing Clean Room run and steers it back to missed gates without expanding declared scope.
- argument-hint
- [ "existing task-manifest.json or artifact roots" ]
- disable-model-invocation
- true
03 SKILL.md 正文
Clean Room Refocus
Refocus realigns the current run to the declared scope, controller policy, artifact schemas, and clean-room boundary.
Refocus does not optimize, expand, or reinterpret the task. It does not invent new requirements or add behavior beyond preflight-goal.json, task-manifest.json, clean-run-context.json, ledgers, implementation plan/report, QC, and abstract delta tickets.
Use the canonical clean-room skill workflow and references in this plugin. Preserve the same clean-room boundary, role separation, artifact schemas, leakage rules, implementation-root rules, and hook expectations.
Audit
Compare current artifacts to the canonical gate checklist:
- Scope gate recorded authorization, roots, boundaries, prohibited actions, evidence handling, selected target profile, Agent 0-4 pipeline, and Agent 1.5 sanitizer role for new runs.
- Preflight goal exists for new runs, validates against
preflight-goal.schema.json, and is referenced bytask-manifest.jsonwithpreflight_goal_refandpreflight_goal_sha256. - Task manifest records the required
handoff_sequenceand does not skip Stage 0. - Initialization snapshot exists when init preferences were used, and reusable
init-config.jsondrift is reported instead of silently applied. clean-run-context.jsonexists before clean roles run and excludes source roots, visual roots, contaminated roots, source index refs, visual index refs, and ledger paths.clean-run-context.jsonrecords artifact-only coordination: Agent 0 does not directly steer Agent 2, Agent 3, or Agent 4, and clean implementation/polish roles report to Agent 0 only at terminal status.- When context management is enabled, the next role launch can be driven by a fresh
role-session-brief.jsoninside the recorded budgets.controller-status.jsonremains contaminated-side only. - Implementation roots are recorded, separated, and not source-derived.
- Source index or visual-index fallback preflight exists when required and remains contaminated-only.
- Decomposition produced neutral
task-manifest.jsonunits. - Contaminated analysis wrote draft behavior specs without implementation code.
- Agent 1.5 sanitization ran before handoff, or older artifacts clearly predate the Agent 1.5 gate and require review before reuse.
- Handoff package excludes full
task-manifest.json,source-index.json,visual-index.json, raw screenshots, source or visual paths, raw diffs, copied comments, copied visible words, private identifiers, exact UI palettes/layouts/iconography, source-shaped pseudocode, and contaminated ledgers. - Agent 2 produced clean-only
implementation-plan.jsonwhen the run reached that gate. - Agent 3 produced
implementation-report.jsonwhen the run reached implementation. - Non-terminal Agent 3 implementation reports were not used as Agent 0 feedback or guidance.
- QC produced
qc-report.jsonwith schema, leakage, coverage, and abstract delta ticket status when the run reached that gate. - Agent 4 produced
polish-report.jsonwhen the run reached final polish review. - Contaminated verification returned only abstract delta tickets.
Validate schemas and handoff hashes before trusting the artifacts. Use source-index.json or visual-index.json only on the contaminated side and only when referenced by the task manifest.
Findings
Emit missed-gate findings only:
- Missing required artifact.
- Invalid schema.
- Failed or missing leakage review.
- Invalid or stale handoff hash.
- Boundary violation or unproven root separation.
- Stale handoff compared with latest QC or ledger state.
- Stale implementation plan compared with latest clean handoff.
- Stale implementation report compared with latest implementation plan.
- Controller policy not preserved.
- Missing, invalid, or drifted preflight goal.
- Noncanonical manifests, reports, ledgers, or manual result summaries used as completion evidence. Mark these
not verifiedunlessclean-room-skill run --dry-run(ornpx clean-room-skill@latest run --dry-runif the binary is not available) succeeds against the canonicaltask-manifest.json. - Missing public-surface inventory parity: required public commands, APIs, config keys, protocol entries, or user-visible behaviors listed in approved specs are not mapped through behavior spec tests, implementation-plan
public_contract_refs, terminal implementation reports, and coverage-ledgerpublic_surface_coverage.
Do not suggest speculative improvements. Do not change source scope, target profile, public API, or implementation plan. If the user asks to add scope, stop and route to a new scope gate instead of silently expanding the run.
Output
Return a bounded next-action plan containing:
- Current verified gate.
- Missed gate findings.
- One to three process corrections, ordered by safety.
- The single next action the controller should take.
If root separation, authorization, or clean/contaminated boundary cannot be proven, the single next action is to stop and repair that gate.
04 同仓库更多 Skills
worktable-review-with-annotations
Create, inspect, reply to, update, or resolve situated Worktable annotations. Use when feedback belongs to a Doc section, block, text range, or HTML Doc, or when the user asks to address existing review feedback and coordinate a requested artifact edit.
worktable-find-and-synthesize-context
Find, assess, and synthesize context from Worktable Docs and Records. Use when the user asks what a workspace contains, wants workspace-backed research, needs current facts reconciled across artifacts, or requests a cited synthesis without changing workspace state.
worktable-create-or-update-docs
Create, edit, rename, or maintain narrative Worktable Docs. Use for plans, decisions, briefs, notes, research syntheses, and other durable prose when the user requests a new artifact or a safe update to an existing one.
worktable-create-or-manage-records
Organize repeated structured information as durable Worktable Records. Use for research datasets, inventories, directories, pipelines, trackers, and other collections whose items need identity, fields, filtering, links, validation, or individual updates.
worktable-create-or-manage-interactive-html
Create or improve durable visual and interactive artifacts in Worktable. Use for dashboards, trackers, calculators, mini apps, slide presentations, and other requests that benefit from a polished HTML Doc.
worktable-collaborate-threads
Read, answer, and hand off messages through durable Worktable threads with users or other authenticated participants. Use when the user asks to check addressed messages, reply to a participant, continue a thread, request another agent's help, wait briefly for a response, or watch for comments and messages sent to the agent.