Mathieu Benoit
5fb49da0f2
Le dépôt n'avait aucune règle sur le commentaire de code : le seul modèle d'écriture était le corps de commit, et le gabarit y renvoyait le raisonnement qui n'y tenait pas. Le récit s'écoulait dans les sources, avec les noms qu'il portait. L'épreuve tient en une phrase : le code est le sujet, au présent de ce qu'il fait. Une phrase dont le sujet est un incident, une machine, une date ou une personne part vers tasks/, non versionné ; le mode de défaillance que le code empêche reste. Rien d'identifiant hors de private/, et seulement sur un dépôt privé. --- EN --- The repository had no rule at all about code comments: the only writing model available was the commit body, and the template sent the reasoning that did not fit there into a comment. The story flowed into the sources, carrying the names it named. The test fits in one sentence: the code is the subject, in the present of what it does. A sentence whose subject is an incident, a machine, a date or a person goes to tasks/, unversioned; the failure mode the code prevents stays. Nothing identifying outside private/, and only on a private repository. Assisted-by: Claude Opus 5
52 lines
2.3 KiB
Markdown
52 lines
2.3 KiB
Markdown
# Workflow Orchestration
|
|
|
|
## 1. Plan Mode Default
|
|
- Enter plan mode for ANY non-trivial task (3+ steps or architectural decisions)
|
|
- If something goes sideways, STOP and re-plan immediately
|
|
- Don't keep pushing.
|
|
- Use plan mode for verification steps, not just building
|
|
- Write detailed specs upfront to reduce ambiguity
|
|
|
|
## 2. Subagent Strategy
|
|
- Use subagents liberally to keep the main context window clean.
|
|
- Offload research, exploration, and parallel analysis to subagents
|
|
- For complex problems, throw more compute at it via subagents
|
|
- One task per subagent for focused execution
|
|
|
|
## 3. Self-Improvement Loop
|
|
- Capture Lessons: Update AGENT.md with any change in approach the user has asked you to make.
|
|
- Write rules for yourself that prevent the same mistake
|
|
- Ruthlessly iterate on these lessons until the mistake rate drops.
|
|
- Review lessons at session start for relevant project
|
|
|
|
## 4. Verification Before Done
|
|
- Never mark a task complete without proving it works
|
|
- Diff behavior between main and your changes when relevant
|
|
- Ask yourself, "Would a staff engineer approve this?"
|
|
- Run tests, check logs, and demonstrate correctness.
|
|
|
|
## 5. Demand Elegance (Balanced)
|
|
- For non-trivial changes, pause and ask, "Is there a more elegant way?"
|
|
- If a fix feels hacky: "Knowing everything I know now, implement the elegant solution"
|
|
- Skip this for simple, obvious fixes
|
|
- Don't over-engineer.
|
|
- Challenge your own work before presenting it
|
|
|
|
## 6. Autonomous Bug Fixing
|
|
- When given a bug report, just fix it. Don't ask for hand-holding
|
|
- Point at logs, errors, and failing tests.
|
|
- Then resolve them.
|
|
- Zero context switching required from the user
|
|
- Go fix failing CI tests without being told how
|
|
|
|
## Task Management
|
|
|
|
1. **Plan First**: Write a plan to `tasks/todo.md` with checkable items.
|
|
2. **Verify Plan**: Check in before starting implementation.
|
|
3. **Track Progress**: Mark items complete as you go.
|
|
4. **Explain Changes**: High-level summary at each step.
|
|
5. **Document Results**: Add review section to `tasks/todo.md`. L'enquête,
|
|
les mesures datées, les impasses et les traces d'exécution restent LÀ.
|
|
`tasks/` n'est pas versionné : il porte ce que ni le code ni le commit ne
|
|
doivent porter. Ne les fais pas remonter.
|
|
6. **Capture Lessons**: Update `tasks/lessons.md` after corrections
|