Audits a single agent file for content that overlaps with the skills it explicitly references — verbatim duplication, conceptual restatement, inlined procedures that should defer to a skill, and duplicated examples/templates. Read-only; emits a severity-grouped report with before/after patch suggestions. Invoke with…
Trims the description frontmatter field of a single skill or agent markdown file down to its essential clauses — what it does, optional "Use when …" trigger, and optional Invoke with: signature. Idempotent; read-modify-write only. Invoke with: @description-trimmer.
Assesses a SKILL.md file for correctness and alignment with the official Claude Code skills documentation (https://code.claude.com/docs/en/skills). Reports issues by severity with citations from the spec. Invoke with: @skill-auditor.
Reads a pattern file and creates a SKILL.md in the specified plugin. Accepts optional metadata overrides via a Provided Metadata block. Invoke with: @skill-from-pattern.
Discovers pattern files in a folder, interviews the user per-pattern to collect metadata, then spawns skill-from-pattern agents to convert each selected pattern into a skill. Invoke with: /convert-patterns.
Discovers every agent and skill inside a plugin directory and dispatches the description-trimmer agent on each one to shrink overly verbose frontmatter descriptions. Use when the harness warns about large cumulative agent or skill descriptions, or when bulk-cleaning a plugin's metadata. Invoke with: /trim-descriptions.
Scaffold a spec-driven microservice repo with a deterministic script: uv src layout, ruff + ty, Makefile, Docker/compose/Liquibase, the keystone containers/settings/entrypoint shells, and the vendored spec-core runtime modules — then initialize every spec layer.
★not rated 5 13d agoA
tokens not measured
originalApache-2.0
Shared base conventions, agents, and reference modules for the spec-driven-development plugins (cross-plugin naming, layout, target-location resolution, and the copied runtime-contract modules).
★not rated 5 13d agoA
tokens not measured
originalApache-2.0
Bootstrap an LLM-maintained knowledge base via a prism-driven interview: start from sample raw inputs (facts), derive the right page types, kinds, and relationships (concepts & categories) with the ontology-prism / KEOPS method, then scaffold the schema, conventions, and operating skills for the new wiki — and dogfood.
★not rated 2 2mo agoA
tokens not measured
At most 3 mods per repository are shown here, and a mod shipped inside a plugin is left to that plugin's page — the rest are on their repository pages: