From agent-harness
Produce an implementation plan before non-trivial code changes — scope, key design decisions, open ambiguities, and which decisions need developer approval vs. which are autonomous. Use before starting any non-trivial implementation, when a request could alter a public contract, persisted schema, or externally observable behavior, or when a task needs to be framed as a verifiable, test-first goal.
How this skill is triggered — by the user, by Claude, or both
Slash command
/agent-harness:planningThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Produce a plan before non-trivial implementation begins. Do not execute
Produce a plan before non-trivial implementation begins. Do not execute without explicit developer approval on the plan.
git fetch +
git status -sb or git log HEAD..origin/<default>). A plan built on
stale local state can target code that's already changed upstream.implementation skill's resolution ladder;
state the assumption inline.Once the plan is approved, proceed to the implementation skill. If
execution has to diverge from the approved plan, that gets flagged
explicitly during implementation — never applied silently.
npx claudepluginhub cosmoscalibur/agent-harness --plugin agent-harnessCreates structured, bite-sized implementation plans from specs or requirements before writing code. Useful for breaking down multi-step tasks into testable steps with file structure and task boundaries.