メニュー
Review Brain Dump ticket work using CLI commands only. Use when implementation is finished and findings, fixes, and demo generation need to be tracked without MCP.
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
Update Brain Dump's AI model pricing catalog (DEFAULT_COST_MODELS) when providers release new models or change prices. Use when the user asks to add a new model, refresh provider pricing, or fix ticket cost attribution for a model that shows $0 or wrong costs. Covers Anthropic, OpenAI, Google, Cursor, OpenCode, and open-source providers.
Use this skill when Ralph is working autonomously through Brain Dump backlogs. Covers ticket selection, implementation patterns, and autonomous workflow management.
MANDATORY quality workflow for Brain Dump tickets. Defines the exact MCP tool call sequence every ticket must follow. Load this before starting any ticket work.
Enforces Brain Dump quality workflow for ticket implementation with automatic telemetry
Autonomous backlog processing workflow for Brain Dump using the Universal Quality Workflow. Use when working through multiple tickets autonomously or when asked to process a backlog like Ralph.
MANDATORY quality workflow for Brain Dump tickets. Defines the exact MCP tool call sequence every ticket must follow. Load this before starting any ticket work.
| name | brain-dump-review |
| description | Review Brain Dump ticket work using CLI commands only. Use when implementation is finished and findings, fixes, and demo generation need to be tracked without MCP. |
Use this skill after code changes are implemented and validation has run.
git diff --stat
git diff
brain-dump context --ticket <ticket-id> --pretty
Discover and run this project's validation commands from docs/config. Do not assume pnpm, npm, TypeScript, lint, or test scripts exist. If no automated validation command is discoverable, run a targeted manual smoke check and record that no project validation command was found.
For each issue found:
brain-dump review submit-finding \
--ticket <ticket-id> \
--agent <code-reviewer|silent-failure-hunter|code-simplifier> \
--severity <critical|major|minor|suggestion> \
--category <category> \
--description "<issue description>" \
[--file <path>] \
[--line <line>] \
[--fix "<suggested fix>"] \
--pretty
Capture the returned findingId.
After fixing an issue:
brain-dump review mark-fixed --finding <finding-id> --status fixed --description "<how it was fixed>" --pretty
brain-dump review check-complete --ticket <ticket-id> --pretty
Proceed only when critical and major findings are resolved.
Write a JSON steps file, then run:
brain-dump review generate-demo --ticket <ticket-id> --steps-file <path-to-json> --pretty
critical: broken feature, crash, data loss, or severe security riskmajor: incorrect behavior, missing validation, serious error handling issueminor: maintainability or clarity issue without user-facing breakagesuggestion: optional improvementUse these agent names consistently:
code-reviewersilent-failure-huntercode-simplifierTrack review work in Brain Dump using CLI commands only, then hand off to human review.