Verify before done

Skill IrfanSadiqRahat/constellation/methodology/verify-before-done

Validate artifact against schema, acceptance criteria, and downstream readiness BEFORE marking a task complete.From its SKILL.md

Install

npx -y skills add IrfanSadiqRahat/constellation --skill verify-before-done

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

2 things to look at

no licenseNo license file was found in the repository. Code published without one is not open source by default, so using it at work is a question for whoever answers licensing questions where you are.
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

2.8 KB, 658 tokens by cl100k_base, as published. Nobody here has run it

verify-before-done

The verdict shape

artifact: <id + type>
checks:
  - id: schema
    result: pass | fail
    detail: <if fail, what>
  - id: acceptance_criteria
    items:
      - criterion: <text from upstream spec>
        result: pass | fail
        evidence: <link, test name, screenshot>
  - id: downstream_consumable
    by: <next role>
    result: pass | fail | unknown
    detail: <what next role would need that's missing>
  - id: smell_check
    result: pass | warn
    findings: [<smell>]
overall: pass | revise | escalate

The three checks (in order)

Check 1: schema validity

Mechanical. The artifact's shape matches the declared output type from the role's SKILL.md frontmatter.

Frontmatter present? required keys filled?
JSON / YAML structurally valid?
Enum values within domain?

Fail = revise. No exceptions.

Check 2: acceptance criteria

Per item in the upstream spec (PRD requirements, ApiContract endpoints, EvalSuite scores). Each item gets a pass/fail with evidence.

"Latency p95 < 300ms" → evidence: traces.json showing median + p95
"Accessibility AA" → evidence: axe report, manual NVDA pass
"No hallucination > 5%" → evidence: eval suite output

"Looks good" is not evidence. Verbal assertion is not evidence.

Check 3: downstream readiness

Will the next role be able to start immediately on this artifact, or will they have to come back for clarifications?

Backend-architect's output ready for api-designer?
PRD ready for design-lead?
ThreatModel ready for appsec-engineer?

If the next role would ping the previous role for clarification → fail this check.

Verdict resolution

overall	action
pass	mark task done; emit handoff event
revise	dispatch back to producer with verdict as feedback
escalate	human-in-loop required; surface in queue

Role-aware smell checks

Each role declares smell-checks in its SKILL.md. verify-before-done runs them:

role	example smell
`react-specialist`	`useEffect` chained for state derivation
`postgres-specialist`	`IN (...)` with thousands of values
`api-designer`	200 + `error` field (status lying)
`prompt-engineer`	prompts > 4k tokens without reason
`migration-planner`	`ALTER TABLE ... ADD COLUMN NOT NULL DEFAULT` on large table

Why this skill is opinionated

Three checks — schema, acceptance, downstream-readiness separately
Evidence required, not claims
Role-aware smell-checks — each role's playbook drives quality gates
Downstream-readiness is a first-class check — does the NEXT role have what they need?

What ships with it

Read from the repository

Just SKILL.md. No reference files, no scripts.