菜单
Activates only for explicit validator-aware commit requests such as "commit with validator", "run validator then commit", "run checks before commit", or "skip validator and commit". Excludes plain commit requests.
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
Scans the project and configures checks and reviews for Agent Validator for requests such as "set up validator", "configure checks and reviews", or "initialize validator for this repo".
Diagnoses and explains validator behavior from runtime evidence for requests such as "why did validator fail", "explain validator behavior", "diagnose validator logs", or "what went wrong in the validator run".
Advances the validator execution state baseline without running checks for requests such as "skip validator", "advance validator baseline", or "mark current tree as validated without running checks".
Shows a summary of the most recent validator session for requests such as "show validator status", "summarize last validator run", or "what failed in the last validator session".
Capture noteworthy review violations for the eval framework. Use when validator-run finds review failures — judges violations and saves notable ones to evals/inventory.yml.
Activates only for explicit checks-only validation requests such as "run validator checks", "checks only", "check without reviews", or validation without AI review. Runs checks without AI reviews.
| name | validator-commit |
| description | Activates only for explicit validator-aware commit requests such as "commit with validator", "run validator then commit", "run checks before commit", or "skip validator and commit". Excludes plain commit requests. |
| disable-model-invocation | false |
| allowed-tools | Bash, Task |
Commit with explicit validator validation intent. Runs agent-validate detect first, validates based on intent (full run, checks only, or skip), handles failures, then commits.
Use this skill only when the user explicitly asks to involve Agent Validator in the commit flow, including full validation, checks-only validation, or intentionally skipping validator state as part of the commit.
Do not choose this skill for a plain "commit", "make a commit", or "commit these changes" request that does not mention validator, gauntlet, checks, validation, or skip.
Run agent-validate detect using Bash:
agent-validate detect 2>&1; echo "DETECT_EXIT:$?"
Check the exit code from the DETECT_EXIT: line:
Parse $ARGUMENTS for a validation intent. Do not prompt the user if a clear intent is found.
| ARGUMENTS pattern | Action |
|---|---|
| Contains "check" or "checks" | Invoke /validator-check (Step 3b) |
| Contains "run", "full", or "all gates" | Invoke /validator-run (Step 3a) |
| Contains "skip" | Run agent-validate skip 2>&1 (Step 3c), then go to Step 4 |
| Empty or no clear intent | Present the three choices below to the user, wait for selection |
Evaluate the "check" or "checks" pattern before the generic "run" pattern, so
phrases like "run checks before commit" select /validator-check.
When prompting the user, present these choices:
Then proceed to the step matching the user's selection.
Invoke /validator-run.
/validator-run skill handles fixing and re-running. After that skill completes, ask the user: "Ready to commit?". Proceed to Step 4 only on confirmation.Invoke /validator-check.
/validator-check skill handles fixing and re-running. After that skill completes, ask the user: "Ready to commit?". Proceed to Step 4 only on confirmation.Run:
agent-validate skip 2>&1
Report the command output to the user, then go to Step 4.
Check whether you have a skill for committing git changes available (excluding validator-commit itself to avoid self-invocation).
<type>: <description>), then run git commit -m "<message>"