What problem does it solve?
Generic Spec-Driven Development (SDD) agents come pre-configured with generic rules that do not align with your project's unique coding standards, testing workflows, review criteria, and architecture patterns. This leads to inconsistent code, hallucinated project patterns, and wasted time re-explaining your project's rules to AI assistants every new session.
Core Features & Use Cases
- Project-Specific Agent Binding: Populates the GEN:rules region of all 6 core SDD agents (implementer, tester, reviewer, researcher/planner, verifier, etc.) with your project's actual, cited rules for coding, testing, reviewing, and architecture.
- Auto Rule Discovery/Generation: Automatically uses existing project rules if they are present, or generates new rules from your project's folder structure, industry best practices, and web sources if no rules exist yet.
- Idempotent Updates: Safe to re-run whenever your project's rules are updated, ensuring all agents stay in sync without duplicate or conflicting rule entries, and enforces a ratchet system that only sharpens rules over time (no deletions).
- Use Case: For teams using Claude Code, Cursor, or Antigravity with Spec-Driven Development workflows, this skill eliminates the need to manually configure each AI agent to match your project's standards, ensuring consistent, accurate outputs from the first session.
Quick Start
Use the spec-harness-generate-agents skill to bind all 6 generic SDD agents in your .claude/agents directory to your project's specific coding, testing, and architecture rules, populating each agent's GEN:rules region with accurate, cited checklists.
