agentsclimarketplace

Definition of done

Skill precise-alloy/ritus/skills/definition-of-done

Loaded automatically by execute-task and verify-task for STANDARD and EPIC tasks - hard gates and per-change-type completion checklists. Do not invoke standaloneFrom its SKILL.md

Install
npx -y skills add precise-alloy/ritus --skill definition-of-done

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

4.0 KB, as published. Nobody here has run it

Definition of Done

When to use

Load this skill for all STANDARD and EPIC tasks. All hard gates must pass before reporting done. Per-change-type checklists are additive - check all that apply to the current task.

Hard gates - apply to all changes

  • Project compiles with zero errors
  • Lint passes with zero errors (warnings flagged, non-blocking)
  • No new untyped values without inline comment explaining why (typed languages)
  • No secrets, API keys, or credentials in changed files
  • Scope clean: no files modified outside the task's stated CONTEXT

Per change type

New service method

  • At least one unit test: happy path
  • At least one unit test: primary failure path
  • Error handling follows code-conventions skill pattern (no naked throws, no swallowed errors)

New API endpoint

  • Auth guard tested: 401 without valid token
  • Tenant/ownership scoping tested: 403 for wrong entity
  • Input validation present for required fields
  • Structured error response returned to client (not stack trace)

New background worker / job

  • Dispatch logic tested
  • Error logged on failure (job ID + error + stack)
  • Retry behavior tested if applicable

Database migration

  • Migration file exists and runs cleanly
  • Data integrity verified: row counts and schema shape spot-checked
  • Irreversibility documented if migration cannot be rolled back
  • Queries in migration use parameterized inputs

Shared type / contract change

  • All importing modules compile after the change
  • Contract test passes (or documented why N/A with justification)

UI change

  • Component renders without errors (check for runtime warnings in console output)
  • Responsive breakpoints handled in code (verify CSS/class logic covers mobile + desktop)
  • Accessibility attributes present: aria-label, alt, role where applicable
  • Form inputs have associated labels (no unlabeled inputs)
  • Loading, error, and empty states handled (not just the happy path)
  • No hardcoded pixel values where relative units are expected (rem/em/% vs px)
  • Browser verification required - add to DONE WHEN: Verified in browser at 375px and 1280px (human)

Safety override changes (auth, billing, migrations, tenant isolation, infra config, shared contracts)

  • Full security checklist in security skill completed
  • Audit log entry created for any state changes
  • Classified STANDARD or above - no TRIVIAL/SIMPLE

Soft gates - flag but non-blocking

  • Lint warning count > 5 new: note in commit message
  • Test omitted for specific edge case: document reason in DONE WHEN
  • Browser verification deferred: document reason - must be logic-only or config-only

Quality thresholds (apply to new or substantially rewritten code only)

  • No single function with cyclomatic complexity > 10 without a comment (count independent branching paths: if/else, switch cases, loops, logical operators, ternaries, catch blocks)
  • No function > 60 lines without structural justification
  • No file > 400 lines - split at module boundary

Pre-existing violations in unchanged code are not your problem - flag them as tech debt if noticed, but do not refactor outside the task's scope.

Completion signal

Report done only after:

  1. All applicable hard gates above checked
  2. Every DONE WHEN item addressed - implemented and self-checked with the build/test/lint you ran (independent verification of each condition is the verify-task gate, not self-verification)
  3. DOC UPDATE completed per doc trigger matrix in task-generation skill
  4. Output block in execute-task format produced

Handoff

  • Report: the standard applied to the parent skill's work.

Keep looking

Skills are one crate of 627,801. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.