菜单
PR-reviewer's diff-level push-gate review — check this task's diff against its spec, then write the validation_record verdict that gates the push. Loaded by pr-reviewer when reviewing a committed task. Self-contained.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Bro's response when something fails — an MCP tool returns is_error=true (halt + surface), the trajectory-server is unreachable (degraded sqlite3 readonly fallback), or the world model is unavailable (retry, identify the lock holder, reconnect the binding). Loaded reactively on the first failure of a session.
Bro's code-touching flow — verify world model, triage the requirement (reuse vs build), propose branch, write spec, dispatch SWE, verify on return, atomic-close. Loaded on the first code-touching ask of a session.
When bro hits a wall — a task leans on a capability the project plainly lacks — and an existing published skill, MCP toolkit, or plugin would close the gap better than hand-rolled code. Bro names the gap, calls cheatcode_search for ranked candidates, judges which best fits this task and codebase, and recommends it for the Human to approve. Loaded when grabbing an external cheatcode beats grinding the capability out from scratch.
Bro's push-gate orchestration — reaping unsigned commits, spawning pr-reviewer per task, and the all-pass push + PR-create + post-merge cleanup path. Loaded by bro when the push hook blocks or the Human asks for review-before-push.
Bro's PR/MR comment triage — resolve the PR, fetch the comment threads, judge which are task-worthy, and dispatch SWE per ratified group. Loaded by bro when /monitor surfaces PR/MR comments.
Generate a new project-local skill at .claude/skills/<name>/SKILL.md and attach it to existing agents. Loads when the user asks to capture a repeatable behavior — e.g. "create a skill that codifies <our convention>", "teach swe to also <run mypy / use black / etc.>", "make a skill for <reviewing PRs / writing changelogs / etc.>", "we need a checklist when <X happens>". Extends the consuming agent's `skills:` frontmatter array only; agent body stays intact. Always Human-approved.
| name | tmb_review |
| description | PR-reviewer's diff-level push-gate review — check this task's diff against its spec, then write the validation_record verdict that gates the push. Loaded by pr-reviewer when reviewing a committed task. Self-contained. |
| allowed-tools | Bash, mcp__plugin_tmb_trajectory-server__pr_monitor_worktree, mcp__plugin_tmb_trajectory-server__validation_record |
Load context via task_brief(task_id) — spec_body, commit_sha, and the changed dirs' world-model scope.
The parent CC session's main checkout may be on ANY branch. Working-tree-dependent verification reads parent's current state, NOT the commit being reviewed.
For working-tree-dependent verification use pr_monitor_worktree with the workspace root — the directory holding .claude/tmb/trajectory.db; it creates the worktree, runs your command, and removes it atomically.
Sha-based git ops (git show <sha>, git diff <sha>~1..<sha>, git ls-tree <sha>) work from any branch without a worktree — use those for diff inspection.
Diff the commit against its parent and judge the diff itself — cite line numbers for every finding:
## Description says, and nothing the typed files[] doesn't name?## Success Criteria bullet visible in the diff (not just claimed)?## Out of Scope item touched, a doc the change made stale?After producing a verdict, YOU (the pr-reviewer subagent) write the validation_attempts row directly, using the spawn prompt's attempt_n and your subagent_session_id. Which path you take depends on your tool list. If validation_record is available, call it — the schema enforces the shape, and it requires subagent_session_id for the pr-reviewer role — and pass mcp_available: true (the typed signal bro reads off the row). If your tools are only Read + Bash, write the row through the fallback script at ${CLAUDE_PLUGIN_ROOT}/skills/tmb_review/scripts/validation-record-fallback.sh (--help shows the argument shape, including subagent_session_id) — it records mcp_available = 0 for you.