Skip to content

/ido4dev:spec-validate

The spec-validate skill runs pre-ingestion validation on a technical spec artifact, checking format compliance and content quality before feeding it into the ingestion pipeline.

What It Does

/ido4dev:spec-validate path/to/technical-spec.md performs two stages of validation:

Stage 1: Format Compliance

  • Verifies the spec follows the ## Capability: / ### PREFIX-NN: heading pattern
  • Checks all required metadata fields are present (effort, risk, type, ai for tasks; size, risk for capabilities)
  • Validates metadata values are in allowed sets
  • Ensures depends_on references point to existing task IDs
  • Runs circular dependency detection on the dependency graph
  • Verifies success conditions are present for each task

Stage 2: Content Quality

  • Checks task descriptions meet minimum length (>= 200 characters)
  • Assesses description quality: does it contain implementation guidance, not just titles?
  • Checks success conditions are specific and verifiable
  • Flags tasks with ai: human that lack explanation of why human involvement is needed
  • Validates effort/risk consistency (e.g., S-effort task with critical risk is suspicious)

Output

The skill produces a structured validation report:

Spec Validation: Real-time Notification System
Format: 12/12 checks passed
Content: 10/12 checks passed, 2 warnings
Warnings:
- NCO-03 (Routing Engine): Description references "user preferences"
but depends_on doesn't include PRF-01. Missing dependency?
- EML-03 (Bounce Handler): risk=medium but effort=S — verify
webhook integration complexity is genuinely small
Ready for ingestion: YES (with warnings)

When to Use

  • After the technical spec writer produces a spec (as part of /ido4dev:decompose)
  • When manually authoring a technical spec
  • Before running ingest_spec to catch issues early
  • As a quality gate in the decomposition pipeline

Metadata Reference

Task Metadata

FieldValuesRequired
effortS, M, L, XLYes
risklow, medium, high, criticalYes
typefeature, bug, research, infrastructureYes
aifull, assisted, pair, humanYes
depends_onComma-separated IDs, or - for noneYes

Capability Metadata

FieldValuesRequired
sizeS, M, L, XLYes
risklow, medium, high, criticalYes