Mode d’emploi
- Copiez la commande et examinez le paquet avant de l’installer dans votre environnement agent.
- Consultez le document original pour les outils, identifiants et environnements requis.
- Décrivez les entrées et le livrable souhaité ; comparez la sortie aux sources.
npx skills add https://github.com/alejandro-cedeno-10/agentic-sdd-kit --skill sdd-checklistAvant de commencer
- L’installation n’inclut ni abonnement au modèle, ni quota API, ni accès aux fichiers.
- Étoiles et installations reflètent l’activité, pas un test indépendant de qualité ou de sécurité.
Rendu depuis le dépôt source en conservant titres, exemples, code, tableaux, liens et images.
Requirements quality checklist
--strict validation proves a spec is well-FORMED; it doesn't prove it's well-SPECIFIED. This skill generates a tailored checklist and grades the change against it, so gaps in clarity or coverage surface before code — not in review.
When to use
- At the
sdd-feature-flowhuman approval gate, to decide whether to approve the spec. - Before
/opsx:applyor launching implementation on any non-trivial change.
Procedure
- Read the change:
openspec/changes/<id>/proposal.md,design.md,tasks.md, and the
delta spec(s); plus openspec/project.md (the constitution) if present.
- Generate the checklist, tailored to what this change touches. Cover at least:
- Completeness — is every stated requirement addressed by the design AND the tasks?
Are error/empty/edge cases specified, not just the happy path? Are non-goals explicit?
- Clarity — is each requirement testable and unambiguous (SHALL/MUST, concrete values)?
Any "should probably / TBD / etc." that hides a real decision?
- Consistency — do proposal, design, tasks and delta agree? Any contradiction in scope,
data model, or naming? Does anything violate the constitution?
- Testability — does each requirement have a scenario or a task that would fail if the
requirement were unmet? Is the verification path concrete?
- Assumptions — is every load-bearing assumption either confirmed or flagged as an open
question (NOT silently baked in)?
- Grade each item pass / fail / n-a with a one-line reason, and list the fails first.
- Recommend: approve as-is, approve with the listed fixes, or send back for revision.
Output
A compact table (item · verdict · reason), fails first, then a one-line recommendation. Read-only — this skill changes no files; it produces the checklist and the verdict.



