cyberneticlibrary
← library

Scan and categorize edge cases systematically

phase1e-scanworkflowsetup L2★0
jheyworth/bmad-auto ↗
What it does

Scan stories for external unknowns and edge cases with execute-access probing

Best for

Surfacing external unknowns that could block Phase 2 launch through execute-access probing (RUN code, not just read), with optional skeptic verification gated on FAIL only.

Inputs
  • · [object Object]
Outputs
  • · [object Object]
Requires
  • · Bash/read access
  • · Grep
  • · CLI execute access for probing
Preconditions

Epic ID and story paths provided; execute access to bash/read tools; per-hunter FINDINGS_SCHEMA enforced via StructuredOutput

Failure modes
  • · Blind read-only scan misses runtime unknowns (critical finding: scanning must use execute access)
  • · INFO placeholder filler inflates findings count without substance
  • · Verify phase (optional) drops valid findings instead of annotating
  • · Cross-layer flags missing or vague
Trust signals
  • · Six SPIKE categories enforced exhaustively (no one-per-category filler)
  • · Execute-access actually used (cli runs, probing, not just reading)
  • · Verify phase is OPT-IN and FAIL-gated (never drops, only annotates)
  • · Cross-layer flags required for each finding