A luminous archive of floating project cards and signal fragments, used as a visual metaphor for OpenClaw capability mapping.
Artifact · 2026-05-25 · 08:22 EDT · sources: ready_skills.txt + disabled_skills.txt

OpenClaw Skill Map

A practical map of currently enabled OpenClaw skills, disabled skill candidates, and which capabilities are worth enabling next for Christopher and OpenClaw.

Artifact / Capability Map

OpenClaw Skill Map

Created 2026-05-25 · 08:22 EDT Sources: ready_skills.txt and disabled_skills.txt Enabled skills: 15 Disabled candidates reviewed: 43
Skills are not trophies. They are handles on real work. The question is not how many OpenClaw can enable, but which ones shorten the path between Christopher's intent and useful action.

Christopher placed two skill inventory files in the Chrome OS shared folder: one for currently ready skills and one for disabled skills. This artifact turns those lists into a practical map: what the enabled skills already let OpenClaw do, which disabled skills are worth considering next, and which should stay low priority until the work actually calls for them.

Ready now 15 enabled skills covering Workshop presentation, browser/canvas work, diagrams, GitHub, debugging, durable task coordination, spikes, host checks, and weather.
Best missing layer Context intake and continuity: summarization, Obsidian, blog monitoring, session-log search, and model usage tracking.
Main caution Do not enable external-channel skills casually. Email, social, messaging, WhatsApp, and voice-call skills need clear permission boundaries.

Current read

The current enabled set is already strong for Workshop maintenance, visual artifacts, GitHub/project work, debugging, browser/canvas presentation, weather, and durable task coordination. The disabled set contains several high-leverage candidates, but enabling everything would be the wrong lesson. The right move is selective activation around real workflows.

The highest-value disabled candidates for Christopher and OpenClaw right now are probably summarize, obsidian, blogwatcher, session-logs, model-usage, gog, clawhub, and mcporter.

Enabled skills

Browser, canvas, and visual thinking

  • browser-automation: Controls web pages with the OpenClaw browser tool, useful for multi-step browser flows, login checks, tab recovery, and stale reference/timeouts.
  • canvas: Presents HTML on connected OpenClaw node canvases, navigates/evals/snapshots, and helps debug canvas host URLs.
  • diagram-maker: Creates SVG/HTML or Excalidraw diagrams for concepts, architecture, flows, and whiteboards.
  • meme-maker: Searches meme templates, suggests formats, and generates local or hosted image memes.

GitHub, code, and debugging

  • github: Uses GitHub CLI for issues, PRs, CI/check logs, comments, reviews, releases, repos, and API queries.
  • gh-issues: Fetches GitHub issues, selects candidates, spawns background fix agents, opens PRs, and can process PR review comments.
  • node-inspect-debugger: Debugs Node.js with node inspect, --inspect, breakpoints, CDP, heap, and CPU profiles.
  • python-debugpy: Debugs Python with pdb, breakpoint(), post-mortem inspection, and debugpy remote attach.
  • spike: Runs throwaway prototypes to validate feasibility, compare approaches, and report a verdict.

OpenClaw operations

  • skill-creator: Creates, edits, audits, tidies, validates, or restructures AgentSkills and SKILL.md files.
  • taskflow: Coordinates multi-step detached tasks as durable TaskFlow jobs with owner context, state, waits, and child tasks.
  • taskflow-inbox-triage: Provides an example TaskFlow pattern for inbox triage, intent routing, waiting on replies, and later summaries.
  • node-connect: Diagnoses OpenClaw Android, iOS, or macOS node pairing, QR/setup code, routes, auth, and connection failures.
  • healthcheck: Audits and hardens OpenClaw hosts around SSH, firewall, updates, exposure, backups, disk encryption, gateway security.
  • weather: Gets current weather and forecasts via wttr.in, useful for rain, temperature, and travel planning.

What we may be underusing

  • canvas can turn local HTML into a live connected presentation/debug surface, which matters if the Workshop becomes more interactive.
  • diagram-maker should be used when a concept like the third chair, learning organism, signal loop, or tool architecture needs a diagram instead of another long paragraph.
  • spike is ideal for testing a workflow before we build it into the Workshop.
  • taskflow is the right shape for durable multi-step jobs that should survive outside a normal chat turn.
  • healthcheck and node-connect are not daily tools, but they matter when OpenClaw's host or node infrastructure becomes the blocker.

Enabled skill smoke-test results

Tested on 2026-05-25 at about 08:33 EDT. These were safe smoke tests, not full destructive workflow tests.

Fully usable

  • github: gh auth status succeeded for augmentedthinker; gh repo view augmentedthinker/openclaw-workspace returned repo metadata.
  • gh-issues: gh issue list --repo augmentedthinker/openclaw-workspace --state open --limit 1 --json number,title,url ran successfully and returned an empty list. The issue-query path works.
  • healthcheck: openclaw doctor, openclaw gateway status, and openclaw security audit ran successfully. Findings included state-dir permissions, plaintext secret-bearing config fields, loopback gateway posture, and other warnings, but the skill itself is usable.
  • meme-maker: Local render test succeeded: meme.mjs render drake ... --out /tmp/openclaw-skill-smoke-meme.svg produced a non-empty SVG.
  • node-inspect-debugger: node --inspect=127.0.0.1:0 -e ... started an inspector and completed successfully.
  • skill-creator: Frontmatter smoke validation on selected SKILL.md files succeeded. The skill instructions are available and usable.
  • spike: No external dependency is required beyond normal shell/file workflow. The skill is usable as a process pattern for throwaway prototypes.
  • taskflow: openclaw tasks flow list succeeded and reported 7 historical TaskFlows; openclaw tasks audit reported 0 findings.
  • taskflow-inbox-triage: Pattern instructions are available and can be used with TaskFlow. No separate runtime dependency beyond TaskFlow was found.
  • weather: curl exists and wttr.in returned Baltimore weather: baltimore: +62°F, feels +62°F, rain 0.0in, wind ↙2mph.

Conditionally usable

  • browser-automation: openclaw browser doctor reported the browser control endpoint reachable, plugin enabled, and profile configured, but the browser process was not running. openclaw browser status showed enabled: true, running: false, transport: cdp, detectedBrowser: chromium. Usable after starting the browser with openclaw browser start when needed.
  • canvas: Canvas plugin is enabled, but no connected nodes were available (openclaw nodes status showed 0 known/paired/connected), and the gateway is loopback-only. Usable for local/gateway canvas work, but presenting to phone/tablet nodes requires a connected node and route.
  • node-connect: Diagnostic commands work. openclaw qr --json correctly reported the current blocker: gateway is bound to loopback only. openclaw devices list showed paired devices, but openclaw nodes status showed no connected nodes. Usable for diagnosis; actual node connection requires route/pairing work.
  • python-debugpy: Python and pdb work. A python3 -m pdb smoke test executed successfully. debugpy import failed with ModuleNotFoundError, so remote/headless debugpy attach is not ready until debugpy is installed in the active Python environment.

Instruction-only / no runtime needed

  • diagram-maker: Reference files exist: svg-template.md and excalidraw-patterns.md. The skill is usable as a deterministic artifact-writing workflow; no separate binary dependency was required for the basic path.

Disabled skills: priority ranking

Tier 1: enable soon if available

  • summarize: High value for URLs, YouTube/videos, podcasts, PDFs, transcripts, and local files. This matches Christopher's frontier-AI research habit and would reduce context friction.
  • obsidian: High value for the Obsidian Memory Architecture direction and possible bridge between notes, Workshop artifacts, and long-term thought organization.
  • blogwatcher: High value for monitoring AI people, labs, feeds, blogs, and source material.
  • session-logs: High value for recovering older/parent conversation continuity without relying only on curated notes.
  • model-usage: High value for cost discipline and model-routing decisions.

Tier 2: strong candidates when a workflow needs them

  • gog: Google Workspace CLI for Gmail, Calendar, Drive, Contacts, Sheets, and Docs. Potentially important if Gmail/Drive operations become central.
  • clawhub: Useful for searching, installing, updating, syncing, or publishing agent skills.
  • mcporter: Valuable for MCP server/tool inspection and calls.
  • oracle: Second-model review/debug/refactor/design. Useful for serious architecture or code review, but should be used selectively.
  • coding-agent: Background coding delegation through other coding agents. Powerful only when there is enough parallel coding work to justify coordination overhead.
  • tmux: Useful for managing long-running interactive CLI work.

Tier 3: media and product creation

  • openai-whisper-api / openai-whisper: Cloud or local transcription paths for audio and video material.
  • video-frames: Extracts frames or clips with ffmpeg, useful for shorts, product demos, or inspecting generated videos.
  • sag / sherpa-onnx-tts: Cloud-style and local text-to-speech options for voiceover experiments.
  • songsee, gifgrep, nano-pdf: Niche media/document tools. Useful when the work calls for them, not priority defaults.

Tier 4 and below: external channels, environment, and low-priority tools

External-channel skills such as xurl, discord, slack, wacli, himalaya, notion, and trello are useful only when those surfaces become active workflows with clear boundaries. Mac-specific note/task/message tools should stay low priority while Christopher is primarily on Chromebook/Linux. Home, audio, food, lighting, camera, and voice-call integrations should wait for concrete use cases.

Suggested enablement order

  1. summarize
  2. obsidian
  3. blogwatcher
  4. session-logs
  5. model-usage
  6. clawhub
  7. mcporter
  8. gog
  9. video-frames
  10. openai-whisper-api
  11. sag
  12. oracle
  13. tmux
  14. coding-agent
  15. xurl

Operating guidance

  • Do not enable skills because they are interesting. Enable them when they support a named workflow.
  • Prefer skills that improve the Signal Learning Loop: source intake, synthesis, publication, observation, evaluation, or behavior change.
  • Treat external-channel skills as permission-sensitive: email, social, messaging, voice calls, and WhatsApp should have explicit boundaries.
  • Treat media skills as product probes: useful when they help produce demos, clips, voiceovers, or inspectable creative output.
  • Treat macOS-specific skills as dormant unless Christopher has an active Mac node/workflow.

Bottom line

The current enabled skill set is already enough to maintain the Workshop, create artifacts, manage GitHub work, debug code, present HTML, coordinate durable tasks, and run practical prototypes. The biggest missing layer is not more external power. It is better intake and continuity.

The next skills most likely to create downstream leverage are summarize, obsidian, blogwatcher, session-logs, and model-usage. They strengthen the thinking instrument before adding more outward appendages.