By danielbrodie
Institutional-knowledge layer for the AI-skill ecosystem. Documents every agent platform, skill packager, installer, and failure mode — and uses the AI inside your agent to navigate them.
Survey the entire skill ecosystem on this machine — every registry, every installer's lock file, every cluster manifest — and produce a ranked list of findings with safe-action proposals. USE WHEN the user asks "audit my skills", "what's going on with my skill setup", "clean up my skill installs", or mentions confusion about where skills live.
Quick health check on skill-tree's manifest — finds dead references, leaf-not-in-cluster orphans, broken routing tables. Reports validation errors and (if cluster routing is set up) the token savings vs flat-catalog mode. USE WHEN the user says "check my skills", "is anything broken", or sees skill-related errors in a session.
Given a user-reported skill-ecosystem symptom ("my skill X isn't appearing", "I get 'skill not found' for Y", "I shipped a new version but my agents don't see it", "things feel slow / cluttered"), match it to a known failure-mode signature, run the targeted diagnostic, and surface the root cause plus the safe fix command per installer. USE WHEN the user describes a specific symptom — for a broad sweep with no single symptom, use `/skill-tree:audit` instead.
Download a SKILL.md from a GitHub URL into the user's local skill library with security checks (prompt injection, zero-width unicode, path traversal). The fetched skill is sandboxed via `disable-model-invocation: true` so it can't auto-fire until explicitly enabled. USE WHEN the user hands you a GitHub URL pointing at a SKILL.md and wants it pulled in safely.
Provision the right skills into THIS project's .claude/skills/. Reads project signals (CLAUDE.md, package manifest, languages) and the global catalog, then picks 5-10 skills that fit. USE WHEN the user is in a project root and wants per-project skills instead of the global catalog.
Own this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimOwn this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimBased on adoption, maintenance, documentation, and repository signals. Not a security audit or endorsement.
Operate and repair the agent skills you already have installed — across Claude Code, Codex CLI, and Gemini CLI.
Skills accumulate, and across multiple installers they break in ways no single tool surfaces: stale plugin-cache versions, symlink rot, parallel registries with version skew, lock-file drift. skill-tree audits every skill registry on your machine, diagnoses why a given skill won't load, and documents how the skill system actually works on each platform — every load-path and frontmatter claim cited to that platform's official docs. That cross-installer operations layer is the core of the tool, and nothing native or otherwise covers it today.
It also includes a provisioner that narrows a project to a small, committable skill set — but be honest about when you need it. Recent Claude Code handles the basic case natively: the skill prelude is auto-capped (~1% of context, least-invoked descriptions dropped first), and skillOverrides / paths hide or scope individual skills with no copy, manifest, or sync. The provisioner earns its keep where that doesn't reach: large catalogs across multiple installers (native per-skill toggles don't affect plugin skills), and team/CI/multi-machine setups that want a reproducible, hash-pinned skill set committed to the repo. For a single-installer setup under the prelude budget, the native settings are simpler — reach for the provisioner at scale.
This is about progressive disclosure and management — not skill discovery. For discovery, see Anthropic's claude-code-setup or ComposioHQ/awesome-claude-skills.
Early. Mostly used by me. Tested on macOS, not Linux. Issues and PRs welcome.
Operate & repair — the core, and the part with no native substitute:
| Command | What it does |
|---|---|
/skill-tree:audit | Walk every skill registry on this machine. Report what's broken and the fix command per installer. |
/skill-tree:diagnose <symptom> | "My new skill isn't appearing." "I get 'skill not found' for X." Match the symptom to a known failure-mode signature in docs/ecosystem-map.md and point at the fix. |
/skill-tree:check | Health check on the manifest graph. |
Provision & sync — useful at scale (large multi-installer catalogs, teams/CI); for a single-installer setup, native skillOverrides/paths are simpler — see the note above:
| Command | What it does |
|---|---|
/skill-tree:provision | Pick a small set of skills that fit the current project and copy them into <project>/.claude/skills/ as a committable, hash-pinned set. |
/skill-tree:sync | Reconcile project skill copies after upstream library changes. |
/skill-tree:fetch <url> | Download a skill from GitHub with security checks. New skills are sandboxed (disable-model-invocation: true) until you enable them. |
Requires Python 3.11+ and uv.
Claude Code
claude plugin marketplace add danielbrodie/skill-tree
claude plugin install skill-tree@skill-tree
Gemini CLI
gemini extensions install https://github.com/danielbrodie/skill-tree
Codex CLI
# Clone somewhere stable, then register the repo root as a local marketplace
git clone https://github.com/danielbrodie/skill-tree ~/.codex/marketplaces/skill-tree
codex plugin marketplace add ~/.codex/marketplaces/skill-tree
# Enable the plugin in ~/.codex/config.toml
echo '[plugins."skill-tree@danielbrodie"]' >> ~/.codex/config.toml
Verified against codex-cli 0.130.0. Codex doesn't have a plugin install subcommand — registering the marketplace plus the [plugins."..."] block in config.toml is the activation. See docs/codex.md for caveats (no SessionStart hook surface, ${CLAUDE_PLUGIN_ROOT} is Claude-specific).
OpenClaw
openclaw plugins install ./openclaw
<project>/.claude/.skilltree.json records which skills got copied in and why:
{
"version": "1.0",
"sourceLibrary": "~/.claude/skills",
"syncedAt": "...",
"skills": {
"tdd": {
"source": "~/.claude/skills/tdd",
"reason": "Python+pytest repo",
"syncedAt": "...",
"sourceHash": "..."
}
},
"auditLog": [...]
}
Skills are copied, not symlinked, so your project edits won't fight upstream library changes. /skill-tree:sync reconciles drift when you want it.
npx claudepluginhub danielbrodie/skill-tree --plugin skill-treeMemory compression system for Claude Code - persist context across sessions
Ultra-compressed communication mode. Cuts ~75% of tokens while keeping full technical accuracy by speaking like a caveman.
Multi-model consensus engine integrating OpenAI Codex CLI, Gemini CLI, and Claude CLI for collaborative code review and problem-solving.
Curate auto-memory, promote learnings to CLAUDE.md and rules, extract proven patterns into reusable skills.
Editorial "Web Designer" bundle for Claude Code from Antigravity Awesome Skills.
Unified status line for Claude Code with multi-CLI (Claude, Codex, Gemini, z.ai) usage monitoring, context, rate limits, and cost tracking