hl/mp
Agent
The fourth move of the spec-driven loop. Captures the knowledge accumulated during a completed feature, makes it findable, updates the project so future agents apply it, and verifies the learning will surface next time. Embodies the Compound Engineering compound step: each unit of work should make the next one easier…
hl/mp
Agent
Produces a spec in docs/specs/ or determines that no spec is needed. Conducts a structured brainstorm if the brief is fuzzy. Skips if the work does not warrant a spec. Trigger phrases: - "write a spec for" - "draft a spec" - "create a spec for" - "spec out this feature" Context: User has a clear, narrow ask. user…
hl/mp
Agent
One-time setup for adopting the flywheel plugin on a project. Creates docs/specs/ and docs/solutions/, and teaches the project's CLAUDE.md (or AGENTS.md) about the spec-driven workflow so agents working in the project know when to use /fw:draft, /fw:review, /fw:compound. Idempotent — safe to re-run. Trigger phrases: …
hl/mp
Agent
Reviews stale entries in docs/solutions/ and proposes targeted updates. Checks docs against the current codebase: missing file references, superseded patterns, conventions that have shifted, and overlap with newer solutions. Maintenance tool — keeps the knowledge store trustworthy as the codebase evolves. When the…
hl/mp
Agent
Checks whether the current implementation satisfies the spec. Produces findings that feed back into /plan. Does not suggest implementation approaches. Trigger phrases: - "review my changes against the spec" - "check implementation against spec" - "what's left from the spec" Context: Implementation is in progress and…
hl/mp
Agent
Validates a spec for quality before implementation begins. Catches ambiguity and gaps that would force an agent to make assumptions during implementation. Returns findings; does not rewrite the spec. Trigger phrases: - "review the spec" - "validate the spec" - "check the spec for gaps" Context: User just wrote a spec…