tibursocampos/agent-dev-toolkit / core/skills/sdd-spec
sdd-spec
VerifiedCreate a PRD for a new feature or change (agent PRD .md; language per LANGUAGE.md + preferences under features/). Feeds sdd-plan. Use when creating a spec or invoking /sdd-spec.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 21.9 KB |
| reference.md | markdown | 1.5 KB |
| references/challenge-vagueness.md | markdown | 2.8 KB |
| references/command.md | markdown | 3.4 KB |
| references/english-override.md | markdown | 327 B |
| references/filename-numbering.md | markdown | 668 B |
| references/product-docs-language.md | markdown | 236 B |
| references/quality-checklist.md | markdown | 2.0 KB |
| references/req-tracking.md | markdown | 1.6 KB |
| references/selective-retrieval.md | markdown | 400 B |
| references/storage-gitignore.md | markdown | 954 B |
| references/template-usage.md | markdown | 720 B |
| references/validate-change.md | markdown | 475 B |
| references/validate-prd.md | markdown | 873 B |
references/challenge-vagueness.md
View on GitHub## Challenge vagueness + product depth (before Write)
Reject or rewrite acceptance / REQ text that cannot be verified. Examples of **forbidden** phrasing: "works correctly", "as expected", "properly", "funciona corretamente", "como esperado", "de forma adequada".
Every CA must state an **observable** outcome (test, script exit code, checklist item, or measurable UI/API result). Every functional REQ maps to ≥1 CA.
### Product depth challenge (FEATURE / STORY / PRD)
Before drafting the PRD, challenge Prior context and operator answers for **depth**, not only wording. Lazy-load norms only at this step (do not preload):
| Artifact | Must challenge / require | Norm / template |
|----------|--------------------------|-----------------|
| FEATURE (sibling) | Problem, Goals, Non-goals, Evidence (omit > fabricate); story table Rationale + Product intent | `feature-altitude.md`, `product-evidence-lite.md`, `templates/features/FEATURE.md` |
| STORY (sibling) | Objective; Who/Job/Outcome when US; OOS; AC budget **happy + rule/edge + failure** with observable Then | `gherkin-budget.md`, `invest-and-story-quality.md`, `templates/features/story/STORY.md` |
| PRD (this Write) | Success **metrics** (§1.3); **MoSCoW** (§4.3); open questions with **Severity** (§5.1) when unknowns remain; when a material decision applies, verify the actual REQ/CA wording reflects it, not only its ID/link | `clarify-depth.md`, `templates/sdd/PRD.md` |
**What / not how:** depth challenges ask for observable outcomes, beneficiaries, and success signals — **never** implementation code, class design, or step-by-step how. Identifiers (types, APIs, paths) stay English when cited.
For decision-backed PRD content, use `skills/_shared/sdd-artifacts/SEMANTIC-SUFFICIENCY.md` rather than duplicating its criteria. Cover rules, failures, compatibility, preservation, OOS, and evidence limits only when material and applicable. A generic acceptance or missing material behavior/failure is a focused gap question, not an agent-authored `agreed` requirement. Keep superseded decisions historical; an unresolved contradiction makes the affected REQ/CA insufficient until the existing decision owner reconciles it.
If FEATURE/STORY siblings are thin: ask all necessary gap questions, in blocks of at most 5, targeting the missing depth fields; do **not** invent Evidence (`product-evidence-lite.md`). If required siblings are missing under FEATURE `needs_*` / brownfield: **`orchestrated`** → STOP per SKILL; **`direct`** → ask (create inline / proceed / optional orchestrate) — do not hard-block.
Anti-task-shatter titles (verb+file/class/script) in Prior context → keep as PLAN/refine altitude; do not promote into Valuable US framing without outcome rewrite (`anti-task-shatter.md` — load only when titles look task-shaped).
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis. | Stars: 471k | Updated | Claude CodeCodex | Verified | Install |
| diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness. | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is available | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |