メニュー
Root cause analysis and bug fix procedure. Investigates bugs, documents the RCA, and creates an implementation prompt for the fix. Use after /investigate or /finding.
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/fluxforgeai/wrought-plugin --skill rca-bugfix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。
Produce an adversarially-verified, source-cited, full-stack, interactive HTML decision brief for informed go/no-go sign-off at a human decision boundary. MVP mode: pivot-review. Extends /ux-design; consumes /research and /design; feeds /design on GO.
Deep multi-agent code review with algorithmic complexity analysis, data structure review, paradigm enforcement, and efficiency analysis. Complements /simplify with deeper dimensions.
End session and create all handoff documentation
SOC 職業分類に基づく
SKILL.md を表示中
| name | rca-bugfix |
| description | Root cause analysis and bug fix procedure. Investigates bugs, documents the RCA, and creates an implementation prompt for the fix. Use after /investigate or /finding. |
| disable-model-invocation | false |
| argument-hint | [finding-ref or description] |
| effort | xhigh |
| wrought | {"version":"1.0","tools":{"capabilities":["read_file","search_content","find_files","write_file","edit_file","run_command"]},"platforms":{"claude-code":{"disable-model-invocation":false}},"agent":{"role":"RCA Engineer","expertise":["root cause analysis","bug fixing","implementation planning"],"non_goals":["architectural redesign","feature development"]},"execution":{"default_mode":"hypothesis-rca","optional_modes":["self-refine"],"max_iterations":12,"max_refine":1,"stop_conditions":["RCA document written to docs/RCAs/","Implementation prompt written to docs/prompts/","User instructed to stop"]},"output":{"format":"markdown","template":"docs/RCAs/{YYYY-MM-DD_HHMM}_{name}.md","required_sections":["Summary","Root Cause","Evidence","Fix","Verification"],"prohibited_content":["credentials","secrets"]},"pipeline":{"track":"reactive","standalone":false,"prerequisites":["docs/investigations/*.md","docs/findings/*.md"],"produces":["docs/RCAs/*.md","docs/prompts/*.md"],"suggested_next":["plan"]}} |
Trigger: Use /rca-bugfix {issue description} when investigating and fixing bugs that require root cause analysis.
Examples:
/rca-bugfix health check timeout causing gray indicators on monitor page/rca-bugfix rate limiting still happening despite batch processing implementation/rca-bugfix extractions stuck at 0 records during streamingBefore executing this skill, verify ALL of the following:
docs/investigations/*.md
--force, proceed without validation.*_FINDINGS_TRACKER.md
The user has provided an issue to investigate. Reason thoroughly about this issue — trace the root cause through evidence before proposing a fix.
Issue to investigate: The argument passed after /rca-bugfix
Before starting, check for upstream artifacts: docs/investigations/*.md (from /investigate), docs/findings/*.md (from /finding), or work from the user-provided description. When a finding report provides a known cause, focus on confirming it, defining the fix, and writing the implementation prompt.
ALL files MUST use this exact format: {YYYY-MM-DD_HHMM}_{issue_name}.md
✅ 2026-01-22_1700_stream_download_timeout.md
❌ RCA_STREAM_DOWNLOAD_TIMEOUT_2026-01-22.md — no prefix, date comes FIRST
See context_check.md
docs/RCAs/3e226861When writing RCAs, use this structure:
# Root Cause Analysis: {Issue Title}
**Date**: {YYYY-MM-DD}
**Severity**: {Critical/High/Medium/Low}
**Status**: {Investigating/Identified/Resolved}
## Problem Statement
{Clear description of the issue}
## Symptoms
- {Observable symptom 1}
- {Observable symptom 2}
## Root Cause
{The actual underlying cause}
## Evidence
{Code snippets, logs, or data supporting the root cause}
## Impact
{What was affected and how}
## Resolution
{How to fix it}
## Prevention
{How to prevent this in the future}
## Debug Strategy (optional — include when fix involves testable code)
- **Self-debug**: enabled | disabled
- **Verifier**: {command} (e.g., `uv run pytest tests/ -q`)
- **Max iterations**: {N}
- **Completion criteria**: {what "done" means — e.g., "bug no longer reproducible, all tests pass"}
- **Escape hatch**: After {N} iterations, document blockers and request human review
- **Invoke with**: `/wrought-rca-fix` (activates Stop hook verifier loop)
When writing prompts for /plan mode, use this structure:
# Implementation Prompt: {Issue Title}
**RCA Reference**: docs/RCAs/{YYYY-MM-DD_HHMM}_{issue_name}.md
## Context
{Brief summary of the root cause}
## Goal
{What needs to be implemented/fixed}
## Requirements
1. {Requirement 1}
2. {Requirement 2}
## Files Likely Affected
- {file1.py}
- {file2.tsx}
## Constraints
- {Any constraints or considerations}
## Acceptance Criteria
- [ ] {Criterion 1}
- [ ] {Criterion 2}
---
## Plan Output Instructions
**IMPORTANT**: Before creating the implementation plan, you MUST enter plan mode:
1. Call `EnterPlanMode` to enter plan mode (compresses context and enables read-only exploration)
2. Explore the codebase and design your implementation approach using read-only tools (Read, Grep, Glob)
3. Write the plan to `docs/plans/{YYYY-MM-DD_HHMM}_{issue_name}.md` including:
- Summary of the approach
- Step-by-step implementation tasks
- Files to modify with specific changes
- Testing strategy
- Rollback plan (if applicable)
4. Call `ExitPlanMode` to present the plan for user approval
5. **Wait for user approval** before proceeding to implementation
6. After plan approval, invoke `/wrought-rca-fix` to start the autonomous bugfix loop with test verification.
---
## Completion
After writing the RCA and implementation prompt:
1. Update the Findings Tracker (per protocol above)
2. Tell the user what was produced:
- RCA saved to `docs/RCAs/{name}.md`
- Implementation prompt saved to `docs/prompts/{name}.md`
3. If context is high (>85%), note: "Context is high. Consider `/session-end`, then `/plan` in the next session."
**CRITICAL PIPELINE RULE**: Suggest ONLY the next pipeline step below. Do NOT offer to implement. Do NOT offer to skip `/plan`. Do NOT offer alternatives to the pipeline sequence.
Next step: Run `/plan` with the prompt at `docs/prompts/{name}.md` to create the implementation plan.
4. **STOP** — do NOT proceed to implementation, do NOT offer to implement. Do NOT add commentary suggesting any pipeline step could be skipped or is unnecessary. Await user instructions.