skills/lofi-monk/lofi-gate/lofi-gate-checkpoint

lofi-gate-checkpoint

SKILL.md

LoFi Gate Judge Skill

[!CAUTION] MANDATORY CHECKPOINT. You cannot commit until this skill approves your changes. Skipping this step will result in rejection at the Gate.

You have completed coding and tests are passing. Now you must judge your own work before declaring victory.

Step 1: Gather Evidence

Run this command to see exactly what you changed:

git diff --staged

If nothing is staged, run:

git diff HEAD

Save the output. This is your evidence.

Step 2: Recall Your Mission

Find the original instructions. Check these locations in order:

  1. The user's original request (in conversation history)
  2. implementation_plan.md (if you created one)
  3. task.md (if it exists)

This is your mission.

Step 3: The Anti-Cheat Check

Look at your evidence (the diff). Answer these questions:

Q1: Did you modify any EXISTING test files?

  • Look for changes to files matching: *.test.*, *.spec.*, test_*.py, *_test.go
  • If YES → Did the mission explicitly ask you to modify tests?
    • NO → VERDICT: FAIL - "Modified existing tests without authorization"
    • YES → Continue

Q2: Did you disable, skip, or weaken any tests?

  • Look for: .skip, .only, @pytest.mark.skip, expect(true).toBe(true), commented-out assertions
  • If YES → VERDICT: FAIL - "Weakened test integrity"

Q3: Did you delete test files or test cases?

  • If YES and not authorized → VERDICT: FAIL - "Deleted tests without authorization"

Step 4: The Scope Check

Compare your evidence against your mission:

Q4: Did you stay within scope?

  • Did you only change what was asked?
  • Did you add files or features that weren't requested?
  • If you went outside scope → VERDICT: FAIL - "Scope creep detected"

Q5: Did you complete the mission?

  • Review each requirement in the mission
  • Is each one addressed in your changes?
  • If incomplete → VERDICT: FAIL - "Mission incomplete"

Step 5: Deliver Judgment

If all checks pass:

✅ JUDGMENT: APPROVED
All changes align with the mission. No test manipulation detected.

If any check fails:

❌ JUDGMENT: REJECTED
Reason: [specific failure reason]
Required action: [what to fix]

Step 6: Act on Judgment

  • If APPROVED: You may proceed to commit/push.
  • If REJECTED: You MUST fix the issues before proceeding. Do not ignore this judgment.

Step 7: Log Your Verdict

Record your decision in the official log:

# If Passed
python .agent/skills/lofi-gate-checkpoint/scripts/logger.py --source "CHECKPOINT" --status "PASS" --message "Approved changes."

# If Failed
python .agent/skills/lofi-gate-checkpoint/scripts/logger.py --source "CHECKPOINT" --status "FAIL" --message "Reason for rejection..."
Weekly Installs
4
First Seen
Jan 29, 2026
Installed on
openclaw3
gemini-cli3
github-copilot3
codex3
kimi-cli3
cursor3