メニュー
Load on strike 3, a tangled run, or a contradiction that cannot be localized. Do not load for first failures, ordinary test red/green loops, or known environment outages.
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
Load before implementing anything that runs continuously or per-item at scale — animation and render loops, timers, pollers, per-request middleware, large-collection iteration. Do not load for one-shot scripts, setup code, or paths that run a bounded handful of times.
Load before hand-writing or hand-editing files a toolchain normally generates or owns — Xcode project files, lockfiles, generated schemas or migrations, IDE workspace files. Do not load for ordinary source code.
Load when entering an unfamiliar codebase, or before proposing an architecture-level change. Do not load for tiny edits, known local fixes, post-implementation summaries, or tasks already scoped by a current state-map.md.
Load for cross-session orchestration — dispatching work to an Albion/GLM worker over tmux, polling a worker completion manifest, or reviewing worker output. Do not load for in-session subagent delegation; use delegation instead.
Load when the stop gate blocks twice on the same task. Do not load for ordinary unfinished work, first gate blocks, or reports where no completion claim is being made.
Load when writing a subagent brief for non-trivial dispatched work. Do not load for direct edits, trivial lookups, user-facing summaries, or delegation already covered by an existing brief.
| name | recovery |
| description | Load on strike 3, a tangled run, or a contradiction that cannot be localized. Do not load for first failures, ordinary test red/green loops, or known environment outages. |
Extend charter §3 (escalation) when the next edit would be guesswork. Stop edits until the contradiction is classified.
| Failure class | Symptom | Cheapest discriminating check |
|---|---|---|
| Theory wrong | Patch is coherent but outcome contradicts prediction | Build a minimal reproducer or oracle that isolates the claimed rule. |
| Edit wrong | Theory still fits, but implementation changed the wrong surface | Inspect diff against task.md; trace caller/callee or file ownership. |
| Test wrong | Product behavior is stable, assertion encodes stale or impossible rule | Compare test premise to docs, fixtures, and live behavior. |
| Environment wrong | Same inputs vary by shell, path, service, time, cache, or dependency state | Re-run in a clean env; print versions, paths, env keys, and fixture state. |
counterexamples.jsonl: hypothesis, case, failure, lesson, next_check.task.md, state-map.md, and the latest diff.Use revert as escalation only. Do not use it to escape uncertainty.
| Revert when | Do not revert when |
|---|---|
The tree no longer matches state-map.md and re-deriving is cheaper than repair. | A single file or hunk can be inspected and corrected. |
| Edits cross the agreed scope fence and cannot be separated reliably. | Tests fail but the failure is localized. |
| Generated or bulk changes obscure the causal patch. | The user or another worker has unrelated changes in the tree. |
Before any git state change, pause for user approval unless the user explicitly authorized it.
Attempted:
Evidence shows:
Decision needed:
Ask for one decision or one missing fact. Include the smallest command, file, or artifact that would unblock the run.