メニュー
Activates when PDE files land in .pde/ during a rise-pde-session-multi-agents-v2 (or rise-pde-session). You become the human's review partner — analyzing PDE output, spotting missed intent, suggesting edits, and watching the clock.
デフォルトでは、最初にソースを確認する Prompt が選択されています。直接コマンドに切り替えるか、ローカルコピーをダウンロードすることもできます。
インストールを決める前に、SKILL.md と SkillsMP に表示されている付属ファイルをお読みください。
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。
直接コマンドでは確認用 Prompt が省略されます。実行前にソースを確認してください。
npx skills add https://github.com/jgwill/dotagents --skill pde-review-companionコマンドは1行のまま表示されます。コピー前に横へスクロールして全体を確認してください。
ローカルで確認しますか？SkillsMP が現在取得できるファイルをダウンロードできます。
Export a JamAI atelier session the way songbird-71bbe83b and atelier-jerry-origin-1937aa47 were dropped — run @miadi/hooks-interpreter on a session_id, classify the /tmp scratchpad creations off the 30-day tmp cleanup, write MANIFEST.tsv, and land the tree under /home/gmusic/<meaningful-name>. Use when someone says "export this session", "like songbird", "hooks-interpreter on Jerry's session_id", "reproduce the atelier drop", or points at .agents/scripts/_env.sh. Reusable. First successful run was 1937aa47 on 2026-08-19; the script takes any session_id.
Use when a state machine or board must be modelled, drawn, validated or generated into code on Eury — and when asking "what does the loom run here", "is stateloom up", "quel port", "the canvas is frozen", "the board never moves", "MCP 401", "how do I open the board from ilex". Also use when deciding at which ALTITUDE to model a session — a meaning board a human and an agent can converse with, versus a mechanics board that measures the agent loop — and before generating either. Covers the containerised deployment, its memory ceiling, its boot behaviour, the two-layer rule, and what a transcript can and cannot tell you.
Use inquiry-weave to relate artefacts and episodes (the three-identity weave), author episode-to-episode lineage edges, build the chronicle catalog, and enumerate the Twine story library (MIADI_STORIES_ROOT).
SOC 職業分類に基づく
SKILL.md を表示中
| name | pde-review-companion |
| description | Activates when PDE files land in .pde/ during a rise-pde-session-multi-agents-v2 (or rise-pde-session). You become the human's review partner — analyzing PDE output, spotting missed intent, suggesting edits, and watching the clock. |
| argument-hint | <path-to-pde-file.md> or just the PDE ID |
| user-invokable | true |
| disable-model-invocation | false |
Role: You are the human's analytical partner during the review window of a /rise-pde-session-multi-agents-v2 (or /rise-pde-session). Another agent has decomposed the work and is sleeping. No execution has started yet — the agent will act on whatever is in .pde/ when it wakes. Your job: help the human refine the PDE before the agent wakes and begins working.
TUG.md, __.md, or similar in the project root)pde_decompose (likely 2x) and committed .pde/ filessleep 720).pde/*.md files before the agent wakes and acts on themCritical understanding: Empty output directories (like
deep-search/,orchestration-thesis/) are expected at this stage — they are planned destinations, not missing work. The agent hasn't started yet. You are reviewing the plan, not the results.
.pde/*.md and .pde/*.json filesTUG.md, __.md, or other input files in the project root (the filename may vary per project)Secondary Intents against the original inputAmbiguity Flags — are they real ambiguities or things the human already decided?.pde/<file>.md: ...".pde/*.md files directly, do it.pde/ and issues/ during this window