meta-dev

User-invoked pass that re-derives a project's Claude-native instruction layout — a surface AGENTS.md plus path-scoped .claude/rules/ — as a projection of the installed skill set and the actual codebase. Enumerates every available skill, maps which applies to which path (wiring unwired-but-needed ones, dropping inapplicable ones), grounds every claim against the code, discloses only non-code-recoverable info per path, and drops the inert references/ layer. Read-only toward user-owned notes (roadmaps, scratchpads) — it reports promotion candidates, never rewrites them. Invoke as /meta-dev:revise-agents-md.

Bootstrap a project's instruction files from scratch — a surface AGENTS.md plus path-scoped .claude/rules/ derived from the installed skills and the actual code. Use when the user asks to set up CLAUDE.md/AGENTS.md, bootstrap instruction files or project memory, or initialize the layout for a repo that has none. Not for editing an existing layout's content (update-instructions), tightening one already in place (improve-instructions), or re-deriving the whole layout of a repo that already has one (revise-agents-md).

Raise the quality of a project's instruction files (CLAUDE.md, AGENTS.md, references) WITHOUT changing what they commit to. Use when the user asks to clean up, debloat, or tighten CLAUDE.md or AGENTS.md, audit instruction files, or says their constitution is bloated, messy, or hard to read. Fact or guard changes belong to update-instructions; first-time bootstrap belongs to setup-instructions.

Raise an existing skill's quality against a measurable criterion WITHOUT changing its intended behavior. Use when the user asks to debloat, tighten, shorten, or clean up a skill, fix its formatting or wording, rewrite a description already diagnosed as under-triggering (diagnosing why belongs to review-skill), or run the automated description-optimization loop. Behavior changes belong to update-skill; skills over their size budget belong to restructure-skill.

Reorganize skill architecture — split one skill into several, merge overlapping skills, or move content between SKILL.md, references, and scripts. Use when a skill exceeds its size budgets, when skills overlap or their descriptions collide, when content sits at the wrong loading level, or when the user asks to split, merge, rename, or reorganize skills. Content edits within one healthy skill belong to update-skill or improve-skill.

Read-only audit of a project's instruction files (AGENTS.md, CLAUDE.md, .claude/rules) — never edits them. Use to review, check, audit, or diagnose instruction files, verify every documented claim against the actual codebase, find stale or wrong facts, spot cross-file contradictions or bloat, or get a quality report before deciding what to change. Not for applying fixes (update-instructions / improve-instructions) or first-time bootstrap (setup-instructions).

Entry point and router for ALL work on Agent Skills. Use whenever the user wants anything done to a skill — make, build, edit, extend, tighten, reorganize, audit, or remove one — especially when the operation is unclear, mixed, or spans several, or when the user invokes skill-creation by name.

Change a project's instruction-file content — AGENTS.md, CLAUDE.md, and .claude/rules/ — and re-scope a rule when its area outgrows it. Use for "add/change X in AGENTS.md", "update a rule's skills or facts", "the stack table is out of date", "split the app rule". Not for quality-only cleanup with unchanged facts (improve-instructions), a fresh repo with no layout yet (setup-instructions), or a full re-derivation of the whole layout (revise-agents-md).

Extend or change an EXISTING skill's behavior to meet new requirements. Use when the user asks to add functionality, cover a new case, change what a skill does, or fold new knowledge into a skill — any behavioral edit. Not for quality-only cleanups with unchanged behavior (improve-skill) and not for reorganizing files across skills (restructure-skill).

Create a brand-new Claude Code subagent (registered agent .md) from scratch. Use when the user asks to create, add, or write an agent or subagent, or needs a dedicated worker, reviewer, or judge agent for a task. Not for editing an existing agent — route to update-agent or improve-agent — and not for auditing one, which is review-agent.

Design and author a Claude Code hook for event automation or enforcement. Use when the user asks to add or create a hook, block an action automatically, run something on every edit or session start, or enforce a rule with a hook. Not for auditing or debugging an existing hook — that routes to review-hook.

Create a brand-new Agent Skill from scratch. Use when the user asks to make, build, or scaffold a new skill, wants to turn a workflow, conversation, or body of knowledge into a skill, or needs a capability no existing skill covers. Not for skills that already exist — route edits to update-skill, improve-skill, or restructure-skill via skill-creation.

Raise an existing subagent's quality against a measurable criterion WITHOUT changing its delegation behavior or return contract. Use when the user asks to debloat or tighten an agent, says an agent's prompt is bloated, wants its description or triggering fixed, or asks to audit its tool list. Behavior changes (what the agent does or returns) belong to update-agent.

Safely delete, deprecate, or archive an existing Claude Code subagent. Use when the user asks to remove, delete, retire, or uninstall an agent, says an agent is superseded or unused, or wants to clean up unused agents. Not for plugin-shipped agents living only in the plugin cache — this reports and stops for those.

Safely delete, deprecate, or archive an existing Agent Skill. Use when the user asks to remove, delete, retire, uninstall, or archive a skill, when a skill is superseded, obsolete, or abandoned, or when cleaning up unused skills.

Read-only audit of an existing Claude Code subagent — never modifies it. Use when the user asks to review, check, or audit an agent, asks why an agent isn't getting picked or delegated to, wants its tool list checked, or wants a second opinion on an agent before shipping it. Not for applying fixes — route those to update-agent or improve-agent.

Read-only audit of Claude Code hook configuration and scripts — never modifies them. Use when the user asks to review or audit hooks, asks why a hook didn't fire or didn't block a tool call, or asks whether a hook is safe or correct. Not for authoring or fixing a hook.

Read-only audit of an existing Agent Skill — it never modifies the skill under review. Use when the user asks to review, check, grade, audit, or diagnose a skill, asks why a skill isn't triggering or is misbehaving, wants a quality report before deciding what to do, or wants a second opinion on a skill before shipping or sharing it.

Scaffold a new Claude Code plugin and wire its components. Use when the user asks to create or scaffold a plugin, package skills, agents, or hooks as a plugin, or make a workflow installable/shareable. Not for auditing an existing plugin — route that to validate-plugin.

On-demand currency check: diffs meta-dev's standards and lifecycles against current official Claude Code docs and routes drift through the update lifecycles.

Change an existing agent's behavior — its prompt, tools, model, or preloaded skills — for new requirements. Use for "update/extend agent X", "give the agent tool Y", "change what the reviewer agent does". Not for quality-only cleanup with unchanged behavior (improve-agent), read-only auditing (review-agent), or deleting/deprecating an agent (retire-agent).

Read-only validation and audit of a Claude Code plugin or marketplace — never modifies what it inspects. Use when the user asks to validate a plugin or marketplace, wonders why a plugin won't load, wants the manifest checked, or asks if a plugin is ready to publish.