菜单
Orchestrates the full post-implementation loop: push PR → wait for CI → fix failures → repeat until CI passes or termination rules trigger a pause. This skill should be used when the user says "ship it", "finalize pr", "push and fix CI", "push pr and wait for CI", or invokes "codagent:finalize-pr".
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
Creates a structured implementation task breakdown for a structured change, synthesizing proposal, design, and specs into self-contained per-task files. Use when the tasks artifact is the next step in a change.
How to ask users questions interactively: which tool to use, how to batch questions, and when to ask serially. Follow this skill when asking clarifying questions, requesting confirmation, collecting missing requirements, or pausing for user input. Invoked by other skills rather than directly by the user.
Creates design artifacts through collaborative brainstorming of approaches, architecture, and trade-offs. Use when the user says "design this", "create a design", "brainstorm approaches", or "write a design doc".
Adversarial review of a proposal — challenges the why, what, and how, and suggests alternative approaches. Use when the user says "review proposal", "challenge this proposal", "adversarial review", "poke holes", "stress test the proposal", or wants a second opinion on a proposal before moving to specs/design.
Evaluate whether a software idea is worth building, then write the proposal document. Use when the user wants to assess an idea, says "evaluate", "propose", "is this worth building", or "should we build". If the idea passes evaluation, write the proposal document using the provided template.
Reviews risky/notable assumptions and context gaps surfaced by implementor session reports, fixes high-confidence issues directly, and asks one clarifying question at a time for ambiguous findings. Use when the user says "review assumptions", "audit implementor assumptions", or when invoked by a workflow's assumption-review step.
| name | finalize-pr |
| description | Orchestrates the full post-implementation loop: push PR → wait for CI → fix failures → repeat until CI passes or termination rules trigger a pause. This skill should be used when the user says "ship it", "finalize pr", "push and fix CI", "push pr and wait for CI", or invokes "codagent:finalize-pr". |
Push the PR, wait for CI, fix any failures or review comments, and repeat until the PR is green — or pause when termination rules require human input.
Invoke codagent:push-pr to commit, push, and create or update the pull request.
If push-pr fails (e.g. no remote, auth error), report the error and stop.
Invoke codagent:wait-ci to poll CI status and gather review comments.
The wait-ci skill returns one of four statuses:
passed — CI green, no blocking reviews, no PR commentsfailed — CI failures or CHANGES_REQUESTED reviewscomments — CI green but PR comments need addressingpending — checks still running after timeoutOn passed: Workflow is complete. Report success with the PR URL. Stop.
On pending: Report which checks are still running. Use the appropriate tool for asking the user a question or requesting input to ask whether to wait longer or proceed.
On failed or comments: Record the failure signature (the set of failing CI check names, or comments-only if CI passed but comments exist), then proceed to Step 4.
Check termination rules before attempting a fix:
Max 3 fix cycles. If this would be the 4th fix attempt, pause immediately — show current CI status and use the appropriate tool for asking the user a question or requesting input to ask how to proceed. Do NOT attempt a 4th cycle.
Same failure persists after 2 fix attempts. If the failure signature from Step 3 matches the previous cycle's failure signature AND this is the 2nd consecutive attempt at the same failure, pause immediately — explain the persistent failure in detail and use the appropriate tool for asking the user a question or requesting input to ask for guidance. "Same failure" means the identical CI check name(s) appear as failing across two consecutive wait-ci results after fix attempts.
If neither termination rule applies:
codagent:fix-pr to address CI failures and/or review commentsMaintain a running count and history across cycles:
| Cycle | Failure signature | Action |
|---|---|---|
| 1 | ["lint", "test-unit"] | fix-pr → retry |
| 2 | ["test-unit"] | fix-pr → retry |
| 3 | ["test-unit"] | PAUSE (same failure persisted 2 attempts) |
Another example:
| Cycle | Failure signature | Action |
|---|---|---|
| 1 | ["lint"] | fix-pr → retry |
| 2 | ["test-e2e"] | fix-pr → retry |
| 3 | ["test-e2e"] | PAUSE (same failure persisted 2 attempts, AND max 3 cycles) |
push-pr, wait-ci, fix-pr) is independently invocable for ad-hoc use