agent-isdd

[Internal — use /isdd instead] Turns approved requirements into a testable design.md, grounded in agent-nelly's brief (if available) and research-consolidator's unified codebase research (Phase 2+3).

[Internal — invoked automatically by hooks/commit_audit_gate.py, or on request] Audits skills/, agents/, commands/, hooks/, and every plugin's INTEROP.md for duplicated/contradictory claims, dangling references, and cross-plugin interop drift, auto-fixing high-confidence findings.

Fast, token-efficient independent verification of a design/plan document before implementation — a scoped, tiered alternative to the built-in Plan subagent. Use when a design doc's claims need checking against the actual codebase before committing to it. Trigger on "verify this design", "sanity-check my plan", "review this design doc", "poke holes in this design", or when about to write design.md/tasks.md and wanting independent grounding first. NOT for open-ended "help me design X from scratch" — this only verifies claims in an already-written doc against code.

[Internal — use /isdd instead] Produces EARS-based requirements.md under a hard completion gate — interview mode from scratch, review-and-rewrite from an existing ticket/PRD, or a Fast Track draft-and-self-approve mode for Track: Fast.

Orchestrates the intent spec-driven workflow — EARS requirements through design and tasks — with hard phase gates, a visible breadcrumb/checklist UI, and goal-aware memory. Hands off to separate implementation/review/memory plugins.

[Internal — use /isdd instead] Resolves workflow state, decides start/continue, repairs stale state, scaffolds artifacts, and controls phase transitions through Requirements/Design/Tasks/handoff.