メニュー
Multi-agent crew for spec-driven development. Three agents — Investigator, Builder, Reviewer — work in sequence to audit, build, and verify. Includes automatic repair loops and strict decision levels.
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。
SOC 職業分類に基づく
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/wfprieto/SignalMode --skill signalcrew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。
SKILL.md を表示中
Protected anti-AI writing hygiene rules. Replaces habitual AI language with clean prose.
Signal over noise. Forces the AI to drop filler, keep substance, and respond with precision. Cuts token waste, reduces cost, and makes every response actually useful. LLM-agnostic. Works across all major AI coding agents.
Compress natural language files (CLAUDE.md, AGENTS.md, todos, preferences, project notes) into SignalMode-speak to reduce input tokens on every future session. Compressed version overwrites the original. Human-readable backup saved automatically as FILE.original.md.
| name | signalcrew |
| version | 2.1.0 |
| description | Multi-agent crew for spec-driven development. Three agents — Investigator, Builder, Reviewer — work in sequence to audit, build, and verify. Includes automatic repair loops and strict decision levels. |
| trigger | /signalcrew |
Coordinate three specialized agents to deliver spec-driven work with built-in quality gates. Each agent has a defined role, bounded scope, explicit handoff protocol, and a strict output contract governed by rules/execution.yaml.
Before proceeding with any action, classify it:
Role: Audit the codebase, gather facts, identify gaps, flag risks. Produce a structured brief that the Builder can act on without ambiguity.
Scope: Read-only. The Investigator never modifies files.
Trigger: /signalcrew investigate <task or spec>
Role: Implement against the spec. Use the Investigator's brief as the starting point. No guessing about requirements.
Scope: Implement only what is in the spec. Do not add features or refactor unrelated code.
Trigger: /signalcrew build <task or spec>
Role: Adversarial review of the Builder's output. Find every real problem before it ships. No flattery. No hedging.
Scope: Review only what the Builder produced.
Trigger: /signalcrew review
/signalcrew <task or spec> — runs the agents in a gated sequence:
Every agent handoff and final summary must follow this exact format:
Result: Complete | Partial | Blocked | Failed
Changed: <brief list of files or states>
Verified: <what was tested/proven>
Not verified: <what was not tested>
Blocked by: <blocker if any, or "None">
Next action: <next step or handoff>
/signalcrew off — deactivate crew mode.