agentleFS
Sign inStart free
willianbswillianbs/skills/feature-implementer/SKILL.md

feature-implementer

Skill0 starsChanged 4 months agoMITView on GitHub

Implements approved tasks against specs, ADRs, and plans with minimal diffs and verified commands. Use only after planning/validation (or lite path for chores). Emits IMPL_REPORT. Never redesigns architecture, adds unrelated refactors, or self-certifies without running available checks.

2 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

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

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈6

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 4

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Verify with project commands.

    line 51
  2. Distinctive:

    Stop on ambiguity—don’t guess product intent.

    line 53
  3. Very specific:

    Secure-defaults checklist: when touching auth/data/network/uploads:

    line 60
  4. Distinctive:

    no secrets in code/logs

    line 63
  5. Very specific:

    Update/add tests per TEST_STRATEGY (or minimal regression for lite).

    line 65
  6. Very specific:

    Run discovered lint/test/typecheck commands.

    line 66
  7. Very specific:

    Use for residual risks in IMPL_REPORT (known gaps).

    line 88
  8. Distinctive:

    Never redesign or add new frameworks “while here.”

    line 113
  9. Distinctive:

    Never mark acceptance criteria met without evidence.

    line 115
  10. Distinctive:

    Never expand scope beyond the task without user approval.

    line 116

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “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

Purpose

Ship the approved change correctly, narrowly, and with evidence.

When to Use / When NOT to Use

Use when: tasks are defined and approved; user says implement/build/code.

Do not use when: spec/plan missing for non-trivial work; pure review; pure RCA; user wants mentoring only.

Preconditions

Full path — refuse if missing/Blocked:

  • CONTEXT_PACK (Complete or MostlyComplete)
  • SPEC_STATUS Ready or ReadyWithConditions
  • ADR_COMPLIANCE not Conflict/Blocked
  • PLAN or TASK_GRAPH (at least one)

Lite path (chore/docs/typo): CONTEXT_PACK lite + explicit user ask is enough.

Inputs / Outputs

Inputs: task from TASK_GRAPH, artifacts above, TEST_STRATEGY if present.

Outputs: code changes + IMPL_REPORT

Upstream / Downstream

Upstream: delivery-planner, test-strategy-designer, adr-enforcer, defect-analyst (fixes).

Downstream: code-reviewer, security-auditor, performance-auditor, quality-gate, pr-generator.

Core Principles

  1. Implement exactly what was approved.
  2. Minimal, focused diffs.
  3. Reuse existing patterns.
  4. Secure defaults for auth/data/network touchpoints.
  5. Verify with project commands; never invent “tests passed.”
  6. No drive-by refactors.
  7. Stop on ambiguity—don’t guess product intent.

Process

  1. Confirm task scope + AC_IDs + DoD.
  2. Consume CONTEXT_PACK (do not rediscover the whole repo).
  3. Implement incrementally per task.
  4. Secure-defaults checklist when touching auth/data/network/uploads:
    • authz on every new endpoint/action
    • validate/sanitize inputs
    • no secrets in code/logs
    • least privilege data access
  5. Update/add tests per TEST_STRATEGY (or minimal regression for lite).
  6. Run discovered lint/test/typecheck commands; record results.
  7. Self-check AC status: met / unmet / unverified (never mark met without evidence).
  8. Ready for review only if ACs met or explicitly unverified with reason.

Evidence Requirements

  • commands_run: exact command + exit outcome.
  • changed_paths complete.
  • Unverified ACs listed honestly.

Stop Conditions / Failure Modes

ConditionAction
Missing required artifacts (full path)Refuse; list what’s needed
ADR Conflict discovered mid-flightStop; adr-enforcer
Spec gapStop; spec-validator
Unrelated refactor temptationReject; stay on task
Tests failFix or report; do not claim Ready

Severity + Confidence

Use for residual risks in IMPL_REPORT (known gaps).

Output Contract

## IMPL_REPORT
Task: ...
Changed paths: ...
AC status: AC-1 met | unmet | unverified
Commands run: ...
Deviations: none | ...
Security checklist: n/a | done
Ready for review: yes/no
Decision: Proceed | ProceedWithConditions | Revise | Block

Handoffs

  • code-reviewer — next
  • security-auditor / performance-auditor — if risk paths touched
  • quality-gate — before merge
  • pr-generator — after review/gate

Never

  • Never redesign or add new frameworks “while here.”
  • Never commit secrets.
  • Never mark acceptance criteria met without evidence.
  • Never expand scope beyond the task without user approval.

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.

Check your setup, free