agentleFS
precise-alloyprecise-alloy/ritus/skills/execute-task/SKILL.md

execute-task

Skill1 starsChanged 4 months agoMITView on GitHub

Use when given a task file to implement - reads context, implements steps, runs tests, reports results. TRIGGER - invoke when user says "implement task", "run task", "execute task", "work on task", or provides a task file path (e.g., docs/tasks/.../*.md)

We split this skill into the rules it gives an agent

4 rules. 0 are common, 1 is distinctive, and 3 are very specific to this project.

  • 0

    Common

    Most of the 5 edited versions we read keep these.

  • 1

    Distinctive

    This author's own choices. Worth a look.

  • 3

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Create this TODO verbatim and mark each item done as it completes:

    line 14
  2. Very specific:

    Use the task's stated scope, keep NON-GOALS excluded, satisfy every CONSTRAINTS line, use INTERFACES Consumes signatures verbatim, and …

    line 37
  3. Very specific:

    Read referenced CONTEXT files and docs/tasks/{branch-slug}/exploration.md when present.

    line 40
  4. Very specific:

    Commit types: feat | fix | refactor | test | docs | chore

    line 88

“Common” is measured: kept by at least 6 in 10 of the 5 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

Execute Task

Implement the approved task contract. Complete every STEP and address every DONE WHEN condition before reporting the implementation result.

Process

Create this TODO verbatim and mark each item done as it completes:

TODO:

- [ ] Read the task file using the applicable task shape below.
- [ ] Load CONTEXT and available PRIOR CONTEXT + exploration log; for bugs, search LESSONS and DECISIONS.
- [ ] Load applicable standards before implementation.
- [ ] Apply the relevant ARCHITECTURE checklist when PATTERN is set.
- [ ] Implement the task against current files and its constraints.
- [ ] Run build + tests per DONE WHEN and the exact VERIFY commands with their expected results.
- [ ] Complete DOC UPDATE and record reusable findings in the exploration log.
- [ ] Report the implementation result and each DONE WHEN outcome.

Task contract

ShapeInterpretation
SIMPLEImplement TASK and satisfy DONE WHEN + VERIFY; follow STEPS when present.
STANDARD / EPICRead all sections, including CONSTRAINTS, INTERFACES, NON-GOALS, and DOC UPDATE.
Investigation case fileUse Proposed Fix as STEPS and Regression Test as DONE WHEN.

Use the task's stated scope, keep NON-GOALS excluded, satisfy every CONSTRAINTS line, use INTERFACES Consumes signatures verbatim, and expose Produces exactly as specified.

  • Read referenced CONTEXT files and docs/tasks/{branch-slug}/exploration.md when present. For bug fixes, search docs/LESSONS.md and docs/DECISIONS.md for the affected module. Resolve PATTERN through docs/ARCHITECTURE.md.
  • Confirm referenced existing paths and each STEP's from: block against current code before applying its to:. Execute a valid STEP as written when the only objection is preference.
  • An unclear STEP, missing referenced input, stale block, or demonstrated code contradiction, compile defect, or security flaw ends the attempt with BLOCKED: <concern + evidence + suggested correction>. Cite file:line.
  • An unexpected command status or output ends the attempt with a failure report. Compare every prescribed result, including the intended assertion failure in a red step.
  • Apply DOC UPDATE when present. Append reusable patterns, gotchas, or decisions to the exploration log using skills/ticket-review/templates/exploration.md, creating the log when needed. A SIMPLE task with neither a doc update nor a reusable finding leaves those artifacts untouched.

Standards

Load and apply every matching standard before editing code:

<!-- Keep in sync with verify-task/SKILL.md -->
Task touchesLoad skill
Any code changecode-conventions
New service / endpoint / worker / bug fixtesting-policy
New business logic, new API endpoint, or bug fixtdd
Auth / billing / migration / tenant isolation / infra config / shared contractssecurity
Any STANDARD or EPIC taskdefinition-of-done

Output format

Return one final report after completing the work, or the blocker/failure when progress stops. Address each DONE WHEN condition in the existing summaries, cite file:line for code claims, and summarize command outcomes and failures.

Files changed:
- path - summary

Docs updated:
- path - what changed   (or: none required)

Commit message:
type(scope): subject

- what changed and why
- key invariant enforced (if any)

Breaking: none | <what breaks>
Migration: none | <migration needed>

Commit types: feat | fix | refactor | test | docs | chore

Handoff

  • Report: your implementation report (files changed + commit message), or BLOCKED: <concern + evidence + suggested correction> when a STEP is unclear or provably wrong.
  • TODO update: Implemented → Verify - dispatch verify-task subagent for this task. BLOCKED → Fix plan - correct the STEPS, then dispatch execute-task subagent.

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.

Reviews can't be read right now.

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.