PromptPartner/agentsmith

Autonomous Run prepares and supervises a finite local implementation run after a terminal spec is accepted. It coordinates a maker and checker, with no pushing, merging, deployment, or external-system writes.

Design Review checks a Facet UI component against DESIGN.md before a UI PR is merged. It reviews tokens, states, accessibility, spacing, surfaces, responsiveness, and story/test coverage.

Example Skill shows the minimal structure for a reusable Claude Code skill. Use it as a starting point when you want to capture a repeatable workflow and replace it with your own instructions.

Handoff checkpoints work for recovery when a session is ending, context is tight, or an external gate is near. It writes durable state and a paste-ready kickoff block so the next session can resume without losing details.

Harness Doctor checks whether an Agentsmith harness is installed correctly and healthy. It reports rule, settings, skills, hooks, verification, and leanness issues with a concrete fix for each finding.

Harness Help explains what the Agentsmith harness does, which profile you’re on, the main rules, and the current safety mode. Use it when you need a plain-language starting point or want to know what to do next.

New Feedback logs a harness lesson or post-incident note so the system improves, not just the current bug. Use it after a correction, human intervention, or repeated re-derivation to capture a bounded fix and non-regression check.

New Research Note scaffolds a durable research note in docs/research so findings stay traceable to sources in the repo. Use it when starting a multi-source investigation or capturing sources you want preserved.

Portable Fixture proves that skill identity comes from metadata rather than its installation path. It is used as a fixture for testing portable skill packaging and identity handling.

Release Check walks the Orchard pre-tag checklist so version bumps, migrations, verify, changelog, commit, and tag stay in sync before cutting a release.

Verify answers whether work is shippable or done by running the project’s verification phases and showing evidence. Use it before a commit or PR when you need pass/fail output instead of a claim.

Wayfinder turns a foggy effort into a decision-complete spec and separate implementation-ticket drafts. Use it when the destination is known but scope, dependencies, or key decisions are still unclear.

Writing Rules helps you write or review text an agent will read, such as instructions, handoff notes, prompts, and validation labels. It is used when wording needs to change agent behavior with minimal token cost.