authoring

Plugin authoring in Adminhead/claude-plugins. Commands read from the repo's manifest — today.

Skills (9)

• adr-writing

  Create, review, and maintain Markdown Architecture Decision Records (ADRs), including status changes, deprecation, and supersession. Use when a repo needs a durable record of a significant architectural decision and its rationale ("write an ADR", "record this decision", "supersede ADR 0003", docs/ad

• agent-config-audit

  Security review of Claude Code and agent configuration (CLAUDE.md, settings.json, .mcp.json, hooks, subagent definitions) for injection text, over-broad permissions, secrets, shell-injectable commands, unpinned MCP packages, and fail-open hooks. Read-only and model-driven; runs no scanner. Use when

• agent-teams

  Plan and run a Claude Code agent team, meaning several teammate sessions that share a task list and message each other. Covers when a team beats subagents, team size, one-owner-per-file splits, contract files, messaging rules, plan approval, and shutdown. Use when asked to spawn teammates, paralleli

• config-audit

  Audit the installed Claude Code setup for bloat and rot (skills, plugins, memory, hooks, permissions, MCP servers, caches), measure each item's context cost, and propose Keep, Improve, Update, Retire, or Merge verdicts with evidence. Cleanup is soft-delete only, confirmed one item at a time. Use for

• create-skill

  Write, restructure, or review a Claude Code skill (SKILL.md plus references/ and scripts/) so it triggers on the right requests and stays lean through progressive disclosure. Use for "write a skill for X", "create a SKILL.md", "my skill never triggers", "my skill fires on the wrong requests", portin

• docs-audit

  Audit documentation against the code in read-only mode, then optionally fix it. Finds stale, missing, or contradictory README, setup, API, config, CLI, and changelog docs, ranks findings P0 to P3 with a line you can open for each side, and runs a script that checks documented commands, relative link

• domain-modeling

  Build and sharpen a project's domain model in conversation by challenging terms, resolving fuzzy language, testing edge cases, and writing a GLOSSARY.md as terms settle. Use when discussing codebase terminology, naming a concept, resolving overloaded words like "account" or "order", or creating or e

• manage-agent-instructions

  Create, restructure, prune, or reconcile a repo's agent instruction files (CLAUDE.md, nested CLAUDE.md, .claude/rules/, CLAUDE.local.md, AGENTS.md) from verified repo evidence. Use when setting up or reorganizing project instructions, deciding root vs nested vs path-scoped rules, making CLAUDE.md an

• map-codebase

  Produce an evidence-based map of an existing codebase (stack, structure, architecture, conventions, integrations, testing, concerns) in which every claim traces to a file or command output and unknowns are marked. Use when asked to "map this codebase", "document the architecture", "onboard me to thi