Notre avis

Cette compétence enregistre la preuve du travail accompli dans un journal de session pour assurer une piste d'audit.

Points forts

Crée une piste d'audit persistante
Facile à implémenter avec des modèles prêts
S'intègre avec l'exécution des étapes
Permet une traçabilité simple

Limites

Processus manuel dépendant de la discipline du développeur
Aucune détection automatique des actions
Consignation dans un fichier pouvant poser des problèmes de confidentialité

Quand l'utiliser

Après avoir terminé des actions importantes ou des étapes d'un plan pour maintenir une piste d'audit.

Quand l'éviter

Pour des actions triviales ne nécessitant pas d'enregistrement, ou dans des environnements où la journalisation dans un fichier n'est pas appropriée.

Exemples

Log step completion

Log the completion of step 3: Add user authentication with JWT. Files modified: src/auth.py, tests/test_auth.py. Outcome: success.

Log file change

Log a file change: edited src/utils.py to fix parsing bug. Reason: the parser was failing on empty input.

Log test run

Log test run: command 'pytest tests/' passed all 42 tests with 0 failures.

name: edge-log description: Log proof of completed work to the session log. Use after completing significant actions to maintain an audit trail.

Log Proof

Since Codex CLI doesn't have automatic PostToolUse hooks, this skill manually logs proof of completed work.

When to Use

Call this skill after:

Completing a step in the plan
Making significant file changes
Running tests
Any action that should be recorded for audit

Instructions

1. Gather Proof Information

Collect:

What was done: Brief description of the action
Files modified: List of changed files
Outcome: Success/failure and any relevant output
Timestamp: Current time

2. Append to Session Log

Add an entry to .proof/session_log.jsonl:

{
  "timestamp": "2025-01-15T14:30:00Z",
  "type": "manual_log",
  "action": "Completed step 3: Add user authentication",
  "files": ["src/auth.py", "tests/test_auth.py"],
  "outcome": "success",
  "details": "Added JWT-based auth, all tests pass"
}

3. Create Log Directory if Needed

If .proof/ doesn't exist:

mkdir -p .proof

4. Append Entry

echo '{"timestamp":"...","type":"manual_log",...}' >> .proof/session_log.jsonl

Log Entry Format

{
  "timestamp": "<ISO 8601 timestamp>",
  "type": "manual_log",
  "action": "<what was done>",
  "files": ["<list>", "<of>", "<files>"],
  "outcome": "success | failure | partial",
  "details": "<additional context>",
  "step": <step number if applicable>
}

Quick Log Template

For common actions, use these templates:

Step Completion

{
  "timestamp": "<now>",
  "type": "step_complete",
  "step": <N>,
  "description": "<step description>",
  "proof": "<evidence of completion>"
}

File Change

{
  "timestamp": "<now>",
  "type": "file_change",
  "files": ["<path1>", "<path2>"],
  "action": "create | edit | delete",
  "reason": "<why this change was made>"
}

Test Run

{
  "timestamp": "<now>",
  "type": "test_run",
  "command": "<test command>",
  "result": "pass | fail",
  "details": "<N tests, M failures>"
}

Integration with Other Skills

The $edge-step skill should call $edge-log after completing each step.

Example workflow:

$edge-step executes the work
Work completes successfully
$edge-log records the proof
active_context.yaml is updated with step completion

Why This Matters

Without automatic logging:

There's no audit trail of what happened
Proof claims are unverifiable
State changes are invisible

Manual logging provides:

Accountability
Debugging history
Evidence for self-assessment

Enregistrement des Preuves d'Exécution

Recommandé pour