agentleFS
celestiaorgcelestiaorg/celestia-app/.agents/skills/implement/SKILL.md

implement

Skill421 starsChanged 19 days agoApache-2.0View on GitHub

Implement a celestia-app code change with the safety workflow - triage, align, implement, invariant review loop, verify. Use when the engineer confirms the invariant workflow at the interactive gate or invokes /implement directly.

192 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use setup-ts-deep-modules, ask-matt and 10 more.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • liufeng1976Skill

    setup-ts-deep-modules

    Wire dependency-cruiser into a TypeScript repo so each package is a deep module — implementation hidden in subfolders, reachable only through its entry-point files. User-invoked.

    Also in 48 projects that use this

  • GoogleSkill

    ask-matt

    Ask which skill or flow fits your situation. A router over the skills in this repo.

    Also in 104 projects that use this

  • RoentekSkill

    loop-me

    Grill me about specs for the workflows I want to build, within this workspace.

    Also in 56 projects that use this

  • richard12511Skill

    claude-handoff

    Hand the current conversation off to a fresh background agent that picks up the work immediately.

    Also in 48 projects that use this

  • AWSSkill

    to-questionnaire

    Turn a decision you can't fully answer into a questionnaire for someone else to fill in.

    Also in 100 projects that use this

  • GoogleSkill

    to-tickets

    Break a plan, spec, or the current conversation into a set of tracer-bullet tickets, each declaring its blocking edges, published to the configured tracker — edges as text in one file per ticket locally, or native blocking links on a real tracker.

    Also in 180 projects that use this

  • +2 more from google

The file

We split this skill into the rules it gives an agent

6 rules. 0 are common, 4 are distinctive, and 2 are very specific to this project.

  • 0

    Common

    Most of the 175 edited versions we read keep these.

  • 4

    Distinctive

    This author's own choices. Worth a look.

  • 2

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    When in doubt, treat as the higher tier.

    line 18
  2. Very specific:

    Write an assumptions note to docs/plans/<task>.md: goal, stated assumptions, invariants in play (by INV number), design decisions, open …

    line 24
  3. Distinctive:

    Write the smallest diff that solves the problem.

    line 29
  4. Very specific:

    Run make build after Go changes.

    line 30
  5. Distinctive:

    Stop when a round produces zero confirmed findings, or after three rounds total.

    line 39
  6. Distinctive:

    Run the tests covering the changed behavior.

    line 44

“Common” is measured: kept by at least 6 in 10 of the 175 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

/implement — safe code generation workflow

Implement the requested change by following these phases in order. Read docs/ai/invariants.md before anything else.

Phase 0 — Triage

Determine the risk tier from the files the change will touch:

  • Risky: anything under x/, app/, pkg/, proto/, or multiplexer/.
  • Light: docs, test-only changes, scripts, tooling, .github/.
  • Consensus-critical: risky, and the diff changes a state transition reachable from ABCI — message servers in x/*/keeper, ante decorators, PrepareProposal/ProcessProposal/FinalizeBlock logic, square construction, migrations, upgrade handling.

When in doubt, treat as the higher tier. Light tier: skip Phase 1 and Phase 3 entirely — implement, then run Phase 4. If the diff grows into a risky path mid-task, re-triage and ask the engineer again.

Phase 1 — Align (risky only, hard gate)

  1. Explore the code involved. Answer every question the codebase can answer before asking the engineer.
  2. Interview the engineer about every unclear or under-defined aspect. Give a recommended answer for each question.
  3. Write an assumptions note to docs/plans/<task>.md: goal, stated assumptions, invariants in play (by INV number), design decisions, open questions, and an implementation plan (ordered steps, files to touch, tests to add — as short as the task allows). Never commit this file.
  4. Present the note and wait for explicit confirmation. Do not write code before the engineer confirms.

Phase 2 — Implement

  • Write the smallest diff that solves the problem. Follow existing patterns in the repo.
  • Run make build after Go changes.
  • For bug fixes: write the reproducing test first, watch it fail, then fix.

Phase 3 — Review loop

  1. Get the full diff (git diff plus untracked files) and spawn these agents from .claude/agents/ in parallel on it: determinism-reviewer, adversarial-input-reviewer, consensus-reviewer, compat-reviewer, simplicity-reviewer. Give each the diff and the path to the assumptions note if one exists. If your tool can't load these agents, use each .claude/agents/<name>.md as a review checklist yourself.
  2. Consensus-critical tier: also spawn adversarial-reviewer in the first round. In later rounds, re-run it only on state transitions whose code changed during fixes. The engineer can request it for any change or skip it for a provably trivial one; record a skip in the report.
  3. For every finding, spawn a finding-verifier agent to try to refute it. Discard refuted findings. Escalate unproven findings to the engineer — never fix them on your own.
  4. Fix all confirmed findings, then run the reviewers again on the full current diff, pointing them to what changed during fixes.
  5. Stop when a round produces zero confirmed findings, or after three rounds total. Escalate any still-open findings to the engineer — never silently drop them.

Phase 4 — Verify

  • make lint and make test-short must pass.
  • Run the tests covering the changed behavior; add tests where behavior changed.

Phase 5 — Report

Summarize for the engineer: what changed and why, assumptions honored, findings fixed vs escalated, and a one-line status per invariant (respected or not applicable).

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.