ratatouille

Implement one prd.json task per iteration using tracer-bullet + compound mindset. Triggers on: ratatouille:.

Safety Notice

This listing is imported from skills.sh public index metadata. Review upstream SKILL.md and repository scripts before running.

Copy this and send it to your AI assistant to learn

Install skill "ratatouille" with this command: npx skills add sebastiaanwouters/dotagents/sebastiaanwouters-dotagents-ratatouille

Ratatouille

Implement one task from prd.json per iteration using tracer-bullet methodology and a compounding engineering mindset. Maintain progress.txt across iterations.

Trigger Formats

  • ratatouille: <short task hint>
  • ratatouille: <file path> (e.g., ratatouille: prd.json)

Required Files

  • prd.json (root)
  • progress.txt (root)

Workflow

0) Bootstrap (Always First)

  1. Read progress.txt (if missing, create it).
  2. Read prd.json. If missing, stop and ask user to run mise-en-place.
  3. If this is the first iteration, ask what to do after each iteration and store in progress.txt.

Post-iteration action options:

  • qa (run tests/checks)
  • simplify (refactor for clarity)
  • deslop (remove duplication, tighten docs)
  • review (code-review + fix)
  • commit (atomic commit)
  • none

1) Pick the Next Task

  • If the user provided a task hint, match it to a tasks[].id or tasks[].title.
  • Otherwise, pick the first status: "todo" task.
  • If multiple tasks are equally valid, ask the user to choose.

2) Tracer-Bullet Implementation

For the chosen task: Use a tracer bullet to build the thinnest end-to-end vertical slice that touches every needed layer. This code is not a throwaway prototype—it becomes the foundation you expand after validating the path works. Aim for one happy path, minimal data, and fast feedback before adding depth.

  1. Identify layers involved (UI, API, DB, config, tests).
  2. Build the thinnest end-to-end slice that works.
  3. Validate it (minimal integration test or manual check).
  4. Expand only after the tracer bullet is working.

3) Compound Engineering Mindset

Apply compounding leverage:

  • Prefer reusable abstractions only after the tracer works.
  • Automate or document any non-obvious learnings.
  • Make improvements that reduce future work (tests, scripts, patterns).

4) Update progress.txt

Append a new section:

Iteration: <N>
Task: <id + title>
Status: done|partial|blocked
Decisions:
- ...
Learnings:
- ...
Next:
- ...
Post-Iteration Actions:
- qa|simplify|deslop|review|commit|none

5) Update prd.json

  • Set task status to done or partial.
  • Add notes if blocked.
  • Keep updated date fresh.

6) Run Post-Iteration Actions

Use the actions recorded in progress.txt:

  • qa: run relevant tests/checks
  • simplify: simplify or refactor touched code
  • deslop: remove duplication, tighten docs
  • review: run a code review and fix issues
  • commit: create an atomic commit

7) Recursive Handoff

If tasks remain, re-trigger ratatouille:

  • If the harness supports /handoff or /new, re-issue the same prompt with updated context.
  • Otherwise, ask the user to re-run the same ratatouille: command.

Notes

  • Always start by reading progress.txt.
  • Implement one task per iteration.
  • Keep changes scoped and verified.

Source Transparency

This detail page is rendered from real SKILL.md content. Trust labels are metadata-based hints, not a safety guarantee.

Related Skills

Related by shared tags or category signals.

Automation

flyctl

No summary provided by upstream source.

Repository SourceNeeds Review
Automation

teacher

No summary provided by upstream source.

Repository SourceNeeds Review
Automation

bitwarden

No summary provided by upstream source.

Repository SourceNeeds Review
Automation

chef

No summary provided by upstream source.

Repository SourceNeeds Review