Claude skills
Claude / Skills
82 SKILL.md files in 12 sources, each with its own page.
Where skills load from
Counts per source
| Source | Skills | Loads in a $HOME session |
|---|---|---|
| user | 10 | yes |
| claudeai | 20 | yes |
| orchestrata | 10 | yes |
| tmux-status | 1 | yes |
| turn-navigator | 1 | yes |
| codetools | 7 | conditional |
| engineering | 10 | yes |
| design | 7 | yes |
| productivity | 4 | yes |
| cowork | 2 | yes |
| repo | 7 | no |
| opencode | 3 | never |
All skills
| Skill | Source | Loads in $HOME | Duplicate | Role |
|---|---|---|---|---|
| agent-messaging | user | yes | yes | Talk to kimi (and other agents) directly — open multiple named kimi sessions with persistent context and send them messages, or leave asynchronous ma… |
| canvas | user | yes | yes | Launch a live, interactive architecture canvas and DRAW while explaining — boxes, labeled arrows, notes, numbered flow steps, two-way sync with the u… |
| decision | user | yes | yes | The ADR discipline for EVERY decision — invoke whenever the user or the work mentions a decision, deciding, a ruling, ratification, an ADR, or a one-… |
| kimi-webbridge | user | yes | yes | Kimi WebBridge lets AI control the user's real browser — navigate, click, type, read, screenshot, and interact with any website using the user's actu… |
| notify | user | yes | yes | Send Mohamed a desktop notification via the notify CLI (critical urgency — punches through fullscreen apps). |
| observe-work | user | yes | yes | Never wait blind — watch running work instead of blocking on it. |
| orchestrata-provision | user | yes | yes | Provision or converge a machine to run orchestrata — the deployed Claude config (~/.claude), the six user services, and the Incus graph fleet (terraf… |
| session | user | yes | yes | Track work across sessions via the session CLI (per-project context.json) — begin/end a session, log work/artifacts/decisions, set handoff, manage go… |
| shell-guards | user | yes | yes | What the PreToolUse Bash guards block and how to satisfy them — the write-through-Bash guard (write_text/open-w/sed -i/tee/redirect), the authoritati… |
| test-isolation | user | yes | yes | Give every test an id and reclaim exactly what it created — the isolation pattern that stops tests leaking into shared databases, Redis, and the user… |
| algorithmic-art | claudeai | yes | no | Creating algorithmic art using p5.js with seeded randomness and interactive parameter exploration. |
| built-in-browser | claudeai | yes | no | Read this skill before the first step that uses the built-in browser, the browser pane inside the Claude desktop app (also called the in-app browser,… |
| canvas-design | claudeai | yes | no | Create beautiful visual art in .png and .pdf documents using design philosophy. |
| chrome-browser | claudeai | yes | no | Read this skill before the first step that uses Claude in Chrome, the browser extension whose tools are named mcp__claude-in-chrome__* (also called C… |
| computer-use | claudeai | yes | no | Read this skill before the first step of any request to do something in an app on the person's own computer (Notes, Finder, System Settings, any desk… |
| deep-research | claudeai | yes | no | Use this skill when the user's prompt requires (1) researching a topic across multiple sources, comparing options or alternatives, analyzing trends o… |
| doc-coauthoring | claudeai | yes | no | Guide users through a structured workflow for co-authoring documentation. |
| docs | claudeai | yes | no | docs (editable docs people share and comment on; the default for any document, named as a doc or not - a document, report, proposal, resume, cover le… |
| docx | claudeai | yes | no | Use this skill whenever the user wants to create, read, edit, or manipulate Word documents (.docx) or Word templates (.dotx). |
| google-workspace | claudeai | yes | no | Read this before the first Google Drive, Docs, Sheets or Slides connector call whenever the task creates or changes a Google file. |
| import-memory | claudeai | yes | no | Import a memory export from another AI assistant into Claude's memory — conversationally, additively, and with the content treated as data. |
| internal-comms | claudeai | yes | no | A set of resources to help me write all kinds of internal communications, using the formats that my company likes to use. |
| mcp-builder | claudeai | yes | no | Guide for creating high-quality MCP (Model Context Protocol) servers that enable LLMs to interact with external services through well-designed tools. |
| morning | claudeai | yes | no | Render the user's morning brief as a styled HTML artifact, or set it up as a recurring weekday task. |
| claudeai | yes | no | Use this skill whenever the user wants to do anything with PDF files. | |
| pptx | claudeai | yes | no | Use this skill any time a .pptx or .potx file is involved in any way — as input, output, or both. |
| skill-creator | claudeai | yes | no | Create new skills, modify and improve existing skills, and measure skill performance. |
| theme-factory | claudeai | yes | no | Toolkit for styling artifacts with a theme. |
| web-artifacts-builder | claudeai | yes | no | Suite of tools for creating elaborate, multi-component claude.ai HTML artifacts using modern frontend web technologies (React, Tailwind CSS, shadcn/u… |
| xlsx | claudeai | yes | no | Use this skill any time a spreadsheet file is the primary input or output. |
| agent-messaging | orchestrata | yes | yes | Talk to kimi (and other agents) directly — open multiple named kimi sessions with persistent context and send them messages, or leave asynchronous ma… |
| canvas | orchestrata | yes | yes | Launch a live, interactive architecture canvas and DRAW while explaining — boxes, labeled arrows, notes, numbered flow steps, two-way sync with the u… |
| decision | orchestrata | yes | yes | The ADR discipline for EVERY decision — invoke whenever the user or the work mentions a decision, deciding, a ruling, ratification, an ADR, or a one-… |
| kimi-webbridge | orchestrata | yes | yes | Kimi WebBridge lets AI control the user's real browser — navigate, click, type, read, screenshot, and interact with any website using the user's actu… |
| notify | orchestrata | yes | yes | Send Mohamed a desktop notification via the notify CLI (critical urgency — punches through fullscreen apps). |
| observe-work | orchestrata | yes | yes | Never wait blind — watch running work instead of blocking on it. |
| orchestrata-provision | orchestrata | yes | yes | Provision or converge a machine to run orchestrata — the deployed Claude config (~/.claude), the six user services, and the Incus graph fleet (terraf… |
| session | orchestrata | yes | yes | Track work across sessions via the session CLI (per-project context.json) — begin/end a session, log work/artifacts/decisions, set handoff, manage go… |
| shell-guards | orchestrata | yes | yes | What the PreToolUse Bash guards block and how to satisfy them — the write-through-Bash guard (write_text/open-w/sed -i/tee/redirect), the authoritati… |
| test-isolation | orchestrata | yes | yes | Give every test an id and reclaim exactly what it created — the isolation pattern that stops tests leaking into shared databases, Redis, and the user… |
| tmux-icons | tmux-status | yes | no | Switch tmux status icon themes or customize individual icons for the claude-tmux-status plugin. |
| help | turn-navigator | yes | no | Show turn navigator keybindings |
| blast-radius | codetools | conditional | no | Knowledge graph and affected-set blast radius. |
| deploy | codetools | conditional | no | Release, promote and install codetools with versioned identity. |
| diagram | codetools | conditional | no | Bidirectional code to diagram generation and round-trip verification using codetools code2diagram and mermaid. |
| realize | codetools | conditional | no | The design to code to verify generator. |
| rtm | codetools | conditional | no | Requirements traceability matrix. |
| snapshot | codetools | conditional | no | Byte-exact source snapshot and AST over a checkout. |
| verify | codetools | conditional | no | Run the verification gates with run-once test selection over the blast radius. |
| architecture | engineering | yes | no | Create or evaluate an architecture decision record (ADR). |
| code-review | engineering | yes | no | Review code changes for security, performance, and correctness. |
| debug | engineering | yes | no | Structured debugging session — reproduce, isolate, diagnose, and fix. |
| deploy-checklist | engineering | yes | no | Pre-deployment verification checklist. |
| documentation | engineering | yes | no | Write and maintain technical documentation. |
| incident-response | engineering | yes | no | Run an incident response workflow — triage, communicate, and write postmortem. |
| standup | engineering | yes | no | Generate a standup update from recent activity. |
| system-design | engineering | yes | no | Design systems, services, and architectures. |
| tech-debt | engineering | yes | no | Identify, categorize, and prioritize technical debt. |
| testing-strategy | engineering | yes | no | Design test strategies and test plans. |
| accessibility-review | design | yes | no | Run a WCAG 2.1 AA accessibility audit on a design or page. |
| design-critique | design | yes | no | Get structured design feedback on usability, hierarchy, and consistency. |
| design-handoff | design | yes | no | Generate developer handoff specs from a design. |
| design-system | design | yes | no | Audit, document, or extend your design system. |
| research-synthesis | design | yes | no | Synthesize user research into themes, insights, and recommendations. |
| user-research | design | yes | no | Plan, conduct, and synthesize user research. |
| ux-copy | design | yes | no | Write or review UX copy — microcopy, error messages, empty states, CTAs. |
| memory-management | productivity | yes | no | Two-tier memory system that makes Claude a true workplace collaborator. |
| start | productivity | yes | no | Initialize the productivity system and open the dashboard. |
| task-management | productivity | yes | no | Simple task management using a shared TASKS.md file. |
| update | productivity | yes | no | Sync tasks and refresh memory from your current activity. |
| cowork-plugin-customizer | cowork | yes | no | Customize a Claude Code plugin for a specific organization's tools and workflows. |
| create-cowork-plugin | cowork | yes | no | Guide users through creating a new plugin from scratch in a cowork session. |
| ask-openrouter | repo | no | no | Ask any OpenRouter-hosted model (GPT/Gemini/Kimi/Qwen/DeepSeek/Llama/other Claude tiers/...) via scripts/ask_openrouter.py — for cross-vendor second … |
| ast-toolkit | repo | no | no | Use the AST toolkit (scripts/serialize_codebase_ast.py) to answer codebase-structure questions, refactor, and generate code — instead of grep/sed/man… |
| decision | repo | no | yes | The ADR discipline for EVERY decision — invoke whenever the user or the work mentions a decision, deciding, a ruling, ratification, an ADR, or a one-… |
| knowledge-wiki | repo | no | yes | Read and write the shared knowledge wiki at ~/workspace/knowledge-wiki using the Wiki.js CLI and the wiki-kimi helper. |
| orchgraph-memgraph | repo | no | yes | Build, deploy, and operate the orchgraph Memgraph 3.11 fork with HA clustering and query modules. |
| test-loop | repo | no | no | The fast inner test loop — select ONLY the tests a change can affect (run_tests.py --affected, affected_tests.py, symbol-precise rts.py), observe run… |
| worktrees | repo | no | no | Create, track, and auto-clean git worktrees via scripts/worktrees.py — mint wt- |
| knowledge-wiki | opencode | never | yes | Read and write the shared knowledge wiki at ~/workspace/knowledge-wiki using the Wiki.js CLI and the wiki-kimi helper. |
| orchgraph-memgraph | opencode | never | yes | Build, deploy, and operate the orchgraph Memgraph 3.11 fork with HA clustering and query modules. |
| trace | opencode | never | no | Debug Python test failures and runtime errors using real execution, stack traces, and systematic root cause analysis |
Notes
This index lists every SKILL.md found on this host. The count is 82, not the 83 of the 2026-10-22 inventory: the hermes skill was in the user skills then and was removed by the 2026-10-23 deploy (a copy remains in the deploy backup), so its page was not created.
What loads where
In a session rooted in $HOME Claude Code loads the 10 user skills, the 20 claude.ai skills synced into ~/.claude/skills/synced, the orchestrata plugin's 10, tmux-status's 1, turn-navigator's 1, and the synced cowork plugins (engineering 10, design 7, productivity 4, plugin management 2). That is 65 loadable skills. Three more groups do not load: the 7 codetools skills load only in sessions rooted in /home/mohamed/workspace (the plugin has local scope); the 7 repository skills under .claude/skills of the pre-commit worktree load only in a session that starts there; and the 3 OpenCode skills are never read by Claude Code. The gitlab plugin has no skills (MCP only).
Duplicates
Ten skills exist twice, byte-identical, in ~/.claude/skills and in the orchestrata plugin, so both copies are listed in every session: agent-messaging, canvas, decision, kimi-webbridge, notify, observe-work, orchestrata-provision, session, shell-guards and test-isolation. The decision skill has a third copy in the repository. Edit all copies or none; the repository copy is the source. The knowledge-wiki and orchgraph-memgraph skills exist in a Claude and an OpenCode variant that have diverged. Trigger overlaps worth knowing: decision against engineering-architecture (ADR); engineering-code-review against the built-in code-review skill and the /review command; kimi-webbridge against chrome-browser and claude-in-chrome.
Dead or stale on this host
Dead: kimi-webbridge (the whole ~/.kimi-webbridge directory is missing), built-in-browser, computer-use and import-memory (desktop-app features, unavailable in the CLI). Stale: orchestrata-provision (probes a ~/prod/orchestrata-ultimate that does not exist; prod is /opt/orchestrata), session (still says session sync is Neo4j), test-isolation (Neo4j fixture names). Missing binaries: kimi-session and agent-mail are not on PATH for agent-messaging; qpdf, pdftk and tesseract for pdf; the npm docx and pptxgenjs packages; gh for productivity-update; ct for the codetools skills; wiki for knowledge-wiki. The "what it drives" table on each page shows the live status, regenerated by scripts/gen-claude-pages.py.
Skills are driven by hooks as well as by the model: decision-adr-trigger points to the decision skill, agent_mail_inbox to agent-messaging, and the guards in hooks cite shell-guards and ast-toolkit. See also commands.
Earlier documentation
Skills are on-demand capability packs: a SKILL.md that the agent loads when a task matches the skill's domain. Each page below documents the skill's full text, every CLI command it references (with arguments), and a workflow diagram.
| Skill | Purpose |
|---|---|
| ask-openrouter | Query OpenRouter-hosted models with credential resolution and a redaction contract |
| ast-toolkit | 25-subcommand AST analysis/refactor CLI plus the ast_api.py FastAPI companion |
| knowledge-wiki | Read/write the shared git-backed Wiki.js knowledge wiki |
| orchgraph-memgraph | Build, deploy, and troubleshoot the orchgraph Memgraph fork |
Loading model
Back to Claude configuration.