Agent Skills
💡 What is Agent Skills?
91 AI agent skills · TOON Format · Cross-platform
Agent Skills is a curated collection of 91 AI agent skills for LLM-based development workflows. Built around the jeo orchestration protocol, it provides:
- Unified orchestration across Claude Code, Gemini CLI, OpenAI Codex, and OpenCode
- Plan → Execute → Verify → Cleanup automated pipelines
- Multi-agent team coordination with parallel execution
🚀 Quick Start
Prerequisite: Install skills CLI before running npx skills add.
npm install -g skills
# Send to your LLM agent — it will read and install automatically
curl -s https://raw.githubusercontent.com/akillness/oh-my-skills/main/setup-all-skills-prompt.md
| Platform |
First Command |
| Claude Code |
jeo "task description" or /team "task" |
| Gemini CLI |
/jeo "task description" |
| Codex CLI |
/jeo "task description" |
| OpenCode |
/jeo "task description" |
🏗 Architecture
graph TD
JEO["🎯 JEO\nCore Orchestration"] --> PLAN["📋 PLAN\nooo + plannotator"]
JEO --> EXEC["⚡ EXECUTE\nteam / bmad"]
JEO --> VERIFY["🔍 VERIFY\nagent-browser"]
JEO --> UI["🎨 VERIFY_UI\nagentation"]
JEO --> CLEAN["🧹 CLEANUP\nworktree"]
PLAN --> OMC["omc\nClaude Code"]
PLAN --> OHMG["ohmg\nGemini CLI"]
PLAN --> OMX["omx\nCodex CLI"]
SURVEY["🔭 survey"] -.-> JEO
OOO["🔄 ooo"] -.-> EXEC
AUTORESEARCH["🔬 autoresearch"] -.-> EXEC
🆕 What's New in v2026-04-18
| Change |
Details |
| plannotator: review-packet structural hardening |
Tightened plannotator into a routing-first visual approval gate for concrete plans, markdown specs, and diffs. It now chooses one primary packet (plan-review, diff-review, markdown-review, platform-setup, or troubleshooting), adds references/intake-packets-and-route-outs.md, updates platform/troubleshooting guidance around native-hook vs manual-review reality, expands evals/evals.json with markdown-review and Codex setup truthfulness cases, and syncs compact/setup/manifest wording so discovery surfaces stop overselling it as a generic plan-only or code-review tool. |
| skill-autoresearch: packet-first structural hardening |
Tightened skill-autoresearch into a smaller repo-local ratcheting front door. It now chooses one packet (ratchet-eligibility, benchmark-readiness, charter-freeze, baseline-score, one-change-mutation, support-sync, final-report, or route-out), allows no ratchet justified before mutation churn, adds references/run-packets-and-route-outs.md, expands evals/evals.json, and syncs compact/manifest/setup wording so the repo advertises skill-autoresearch as the local benchmark-and-ratchet layer instead of a generic eval tool. |
| graphify: live-skill promotion + structural hardening |
Promoted graphify from the hidden nested utilities/graphify folder into the live top-level skill catalog, tightened it into a routing-first graph front door, added references/mode-packets-and-route-outs.md plus references/build-and-fallback-recipes.md, added SKILL.toon, refreshed evals/evals.json, and synced README/setup/manifest surfaces so the durable graph lane is actually installable and discoverable. |
| looker-studio-bigquery: packet-first structural hardening |
Tightened looker-studio-bigquery into a smaller reporting front door. It now starts from one packet (dashboard-spec, slow-dashboard, refresh-shape, audience-split, or exec-handoff), adds references/intake-packets-and-route-outs.md, expands evals/evals.json with a Connected Sheets / exec-handoff case, and syncs SKILL.toon / skills.toon / skills.json plus README/setup wording so the repo advertises the BigQuery reporting lane as packet-first instead of a long BI feature tour. |
| web-accessibility: routing-first structural hardening |
Tightened web-accessibility into a smaller routing-first remediation anchor. It now starts from one primary accessibility packet (semantics-structure, keyboard-focus, labels-announcements, visual-perception-reflow, media-alternatives, or routed-navigation-feedback), adds references/intake-packets-and-route-outs.md, expands evals/evals.json with routed-app and responsive-boundary cases, and syncs SKILL.toon / skills.toon / skills.json plus README/setup wording so discovery surfaces stop drifting back to a generic WCAG/ARIA tutorial. |
| marketing-automation: operator-packet ratchet |
Tightened marketing-automation so the broad marketing front door now resolves vague asks into one operating mode, one primary lane, and one reusable operator packet with owner, dependencies/approvals, and proof instead of just a tactic list. Added references/operator-packet-and-proof-stack.md, expanded evals/evals.json with a legal/analytics dependency case, and synced SKILL.toon / manifest wording plus README/setup discovery surfaces so the repo advertises the sharper packet contract truthfully. |
| sprint-retrospective: structural hardening |
Tightened sprint-retrospective into a routing-first PM anchor for sprint retros, milestone postmortems, remote/hybrid facilitation, and dead-action-item recovery across software, product/ops, marketing/GTM, and game-delivery work. The front door now picks one retrospective mode, reviews prior commitments before new actions, keeps action counts brutally small, routes planning/sizing/daily-sync work outward, and adds references/action-review-and-packet-shapes.md, refreshed evals/evals.json, and synced compact/manifest discovery wording. |
| autoresearch: routing-first structural hardening |
Tightened autoresearch into a smaller Karpathy ML search front door. It now chooses one mode (setup readiness, program.md authoring, bounded run loop, results interpretation, or constrained-hardware adaptation), keeps the immutable prepare.py / 300-second / val_bpb contract explicit, adds references/operating-modes-and-route-outs.md, refreshes evals/evals.json, and sharpens route-outs to skill-autoresearch plus app-level eval / observability tools instead of acting like a giant end-to-end explainer. |
🆕 What's New in v2026-04-20
| Change |
Details |
| testing-strategies: gate-truth handoff ratchet |
Tightened testing-strategies around which gate is actually being decided before expanding into layer talk. It now names merge-gate-truth, release-gate-truth, or scheduled-breadth-truth, adds references/gate-truth-and-release-handovers.md, expands evals/evals.json with protected-branch and Steam/build-checklist boundary cases, refreshes SKILL.toon, and syncs README/setup/manifest wording so the skill stops blurring PR blockers with release-only or platform-launch proof. |
| steam-store-launch-ops: packet-first structural hardening |
Tightened steam-store-launch-ops into a smaller packet-first Steam launch router. It now chooses one primary packet (page-promise-audit, wishlist-signal-check, demo-readiness-gate, event-timing-workback, or launch-ops-runbook), adds references/intake-packets-and-route-outs.md, expands evals/evals.json with route-out and broad-GTM boundary cases, refreshes SKILL.toon, and syncs README/setup/manifest wording so the game lane stays Steam-specific instead of drifting into generic marketing or demo-feedback work. |
🆕 What's New in v2026-04-17
| Change |
Details |
| steam-store-launch-ops: bottleneck-router hardening |
Tightened steam-store-launch-ops into a diagnosis-first Steam launch/store router for indie games. It now separates visibility-acquisition, promise-clarity, proof-demo-readiness, timing-hook-fit, and launch-ops-readiness; makes the current hook explicit (coming-soon page, wishlist anomaly, demo decision, Next Fest, or launch window); recommends one intervention and one next artifact instead of a generic marketing dump; and adds references/diagnostic-model.md, references/event-hooks.md, refreshed evals/evals.json, and synced SKILL.toon without adding a duplicate game-marketing skill. |
🆕 What's New in v2026-04-15
| Change |
Details |
| game-performance-profiler: routing-first structural hardening |
Tightened game-performance-profiler into a smaller bottleneck-first profiling front door. It now leans on references/mode-selection-and-route-outs.md plus the existing packet/device/escalation references, keeps one output contract (Game Performance Profiling Brief), adds a build-failure route-out eval case, and sharpens discovery wording around quick packets, benchmark routes, target-device review, and deliberate profiler escalation instead of a giant optimization essay. |
| agent-browser: fresh-session verification rewrite |
Reframed agent-browser from a generic browser CLI guide into the browser-review lane's fresh-session deterministic verification anchor. It now chooses the clean browser lane first, enforces an observe → act → observe loop, makes route-outs explicit to playwriter, agentation, and plannotator, adds references/modes-and-routing.md plus evals/evals.json, and keeps auth reuse bounded instead of silently sliding into a running-browser workflow. |
| agentation: UI annotation router rewrite |
Reframed agentation from a monolithic install/config catalog into the planning-review lane's exact rendered-UI feedback router. It now chooses among copy-paste review, synced watch-loop, self-driving critique, and platform-setup modes; keeps route-outs explicit to agent-browser, playwriter, and plannotator; adds references/modes-and-routing.md, references/platform-setup-and-hooks.md, references/watch-loop-and-self-driving.md, and evals/evals.json; and fixes top-level discovery surfaces so the repo now consistently advertises 89 skills. |
| git-submodule: routing-first structural hardening |
Tightened git-submodule into a smaller operator front door. It still chooses submodule vs subtree/vendoring/package delivery first, but now pushes mode-specific packets into references/mode-packets-and-hosted-constraints.md, adds a hosted-platform boundary for GitHub Pages public-https:// submodule limits, extends eval coverage with that hosted-constraint case, and refreshes discovery wording so recursive bootstrap, pointer updates, detached-HEAD handling, CI auth, and hosted checkout limits stay visible without turning the skill back into a giant command catalog. |
| bmad-idea: pre-planning concept-router rewrite |
Reframed bmad-idea from a legacy BMAD-CIS command/persona catalog into the repository's pre-planning idea router. It now normalizes early-stage product, GTM, consulting, and game packets; chooses one framing mode (problem framing, audience and value framing, concept shaping, game concept framing, or story packaging); produces one reusable concept artifact; routes cleanly to bmad, task-planning, marketing-automation, or bmad-gds; and ships references/operating-modes.md, references/handoff-boundaries.md, references/concept-packet-template.md, and evals/evals.json without increasing the skill count. |
| genkit: packet-first structural hardening |
Tightened genkit into a smaller routing-first backend AI workflow anchor. It now starts from the current packet (new capability, existing route handler, deployed flow quality, runtime/deploy, or comparison/fallback), adds references/intake-packets-and-fallbacks.md, expands the mode selector with comparison-or-fallback, keeps plain SDK / route-handler and durable-workflow fallbacks visible, extends eval coverage with a thin-route boundary case, and syncs SKILL.toon / skills.toon / skills.json wording so discovery surfaces stay aligned with the server-owned workflow contract instead of drifting back to a generic full-stack AI framework pitch. |
| firebase-ai-logic: client-integration support hardening |
Upgraded firebase-ai-logic from a thin companion note into the Firebase lane's app/client integration anchor. It now chooses among direct feature fit, app wiring, production hardening, and escalation-boundary modes; adds references/modes-and-routing.md, references/production-controls.md, references/feature-packets.md, and evals/evals.json; makes route-outs explicit to genkit for backend workflows and firebase-cli for operator tasks; and removes stale setup guidance that previously blurred app SDK work with backend orchestration. |
🆕 What's New in v2026-04-14
| Change |
Details |
| npm-git-install: packet-first hardening |
Tightened npm-git-install into a smaller routing-first package-delivery anchor. It now starts from the packet teams actually have (temporary bridge, shared bridge, private-auth, artifact fallback, workspace inner loop, or durable distribution), adds references/intake-packets-and-route-outs.md, expands eval coverage with a shared internal tooling case, and syncs SKILL.toon / skills.toon / skills.json wording so compact discovery surfaces stop drifting back to stale generic Git-install language. |
| web-design-guidelines: UI-audit rewrite |
Reframed web-design-guidelines from a thin Vercel-rules fetcher into the frontend cluster's broad interface audit anchor for launch-readiness, polish/consistency, flow-friction, heuristic, and rule-overlay reviews. It now classifies findings across hierarchy, clarity, states, responsiveness basics, accessibility basics, and performance/trust signals; adds explicit route-outs to web-accessibility, responsive-design, ui-component-patterns, design-system, and react-best-practices; and ships references/review-modes-and-categories.md, references/handoff-boundaries.md, references/ui-audit-packet-template.md, and evals/evals.json without increasing the skill count. |
| monitoring-observability: packet-first hardening |
Tightened monitoring-observability into a smaller routing-first observability anchor. It now starts from the packet teams actually have (service-health, telemetry-foundation, dashboard/alert audit, data-pipeline trust, or game live-ops visibility), adds references/intake-packets-and-route-outs.md, expands eval coverage with review-audit and deployment-route-out cases, and syncs SKILL.toon / skills.toon / skills.json wording so compact discovery surfaces stop drifting back to the old generic monitoring-setup behavior. |
| performance-optimization: artifact-first hardening |
Tightened performance-optimization into an artifact-first measurement-led tuning anchor. It now starts from the packet teams actually have (trace, Lighthouse/CWV report, query plan, load-test diff, profiler output, or stakeholder report), adds references/intake-packets-and-escalations.md, sharpens route-outs to monitoring-observability, debugging, code-refactoring, testing-strategies, and game-performance-profiler, expands eval coverage for marketing/CWV and CI benchmark cases, and syncs compact/discovery wording so the front door no longer drifts back to stale generic optimization language. |
| code-refactoring: packet-first structural hardening |
Tightened code-refactoring into a smaller routing-first cleanup anchor. It now starts from the cleanup packet teams actually have (local cleanup, fragile legacy area, repeated migration / codemod, or cleanup-heavy diff shaping), adds references/intake-packets-and-route-outs.md, expands eval coverage with a search-first blast-radius route-out, and syncs SKILL.toon / skills.toon / skills.json so compact discovery no longer advertises a stale generic DRY/SOLID design-pattern helper. |
| changelog-maintenance: packet-first hardening |
Tightened changelog-maintenance into a smaller routing-first release-writing anchor. It now chooses one primary mode plus the smallest truthful output packet (single-entry, summary-plus-links, migration-brief, patch-note-brief, or sync-packet), adds references/output-packets-and-channel-handoffs.md, expands eval coverage with a release-notes + migration-window + sync-followups case, and syncs compact/discovery wording so the skill no longer drifts back to a generic changelog / semver template dump. |
🆕 What's New in v2026-04-20
| Change |
Details |
| backend-testing: packet-first structural hardening |
Tightened backend-testing into a packet-first backend test router. It now starts by classifying one backend testing packet (coverage-plan, fixture-and-reset-plan, contract-and-api-checks, flake-stabilization, or execution-lane-split), adds references/intake-packets-and-route-outs.md, expands eval coverage with a contract-protection case, and syncs SKILL.toon / README / setup / manifest wording so the skill stops drifting back to a generic “write some backend tests” catch-all. |
🆕 What's New in v2026-04-19
| Change |
Details |
| omc: routing-first structural hardening |
Tightened omc into a smaller Claude-first orchestration router. It now distinguishes plugin slash skills (/team, /autopilot, /ralph, /ultrawork) from the shell-side omc CLI (omc setup, omc team, omc ask), adds references/intake-packets-and-route-outs.md, refreshes eval coverage for plugin-vs-CLI and recovery cases, updates the detailed docs/omc/README.md guide to stop using stale /omc:* examples, and syncs README/setup/manifest discovery wording so users see OMC as a truthful runtime router instead of a giant command catalog. |
| ui-component-patterns: structural hardening |
Tightened ui-component-patterns into a routing-first reusable-component architecture skill. It now classifies one primary packet (primitive-boundary, slot-anatomy, controlled-ownership, alternate-root-composition, or docs-verification) before suggesting props, adds references/intake-packets-and-route-outs.md, expands eval coverage with alternate-root and Storybook/docs-verification cases, refreshes SKILL.toon / manifest discovery wording, and keeps design-system, web-accessibility, responsive-design, state-management, and react-best-practices as explicit route-outs instead of drifting back to a generic component-best-practices catch-all. |
| responsive-design: structural hardening |
Tightened responsive-design into a routing-first responsive strategy skill that chooses one primary packet (page-layout, component-slot, dense-data, media-behavior, or verification-reflow) before suggesting CSS. The skill now moves packet routing into references/intake-packets-and-route-outs.md, expands eval coverage with a launch-readiness boundary case, refreshes SKILL.toon / manifest discovery surfaces, and keeps ui-component-patterns, web-accessibility, design-system, and web-design-guidelines as explicit route-outs instead of letting responsive work sprawl into a generic frontend catch-all. |
| testing-strategies: structural hardening |
Tightened testing-strategies into a packet-first validation-policy router. It now starts from one policy packet (change-risk, gate-design, flake-cost, release-readiness, or incident-ratchet), adds references/intake-packets-and-route-outs.md, expands eval coverage with release-readiness and accessibility-boundary cases, refreshes SKILL.toon / manifest discovery wording, and keeps backend-testing, debugging, code-review, web-accessibility, and performance-optimization as explicit route-outs instead of drifting back into a giant generic testing manifesto. |
🆕 What's New in v2026-04-13
| Change |
Details |
| responsive-design: layout-adaptation rewrite |
Reframed responsive-design from a long generic CSS example dump into the frontend cluster's mobile-first, container-aware layout adaptation skill. It now classifies viewport-vs-container failures, prioritizes intrinsic layout before breakpoint sprawl, adds explicit route-outs to ui-component-patterns, web-accessibility, design-system, and web-design-guidelines, and ships references/layout-decision-checklist.md, references/handoff-boundaries.md, and evals/evals.json without increasing the skill count. |
🆕 What's New in v2026-04-12
| Change |
Details |
| bmad: packet-first BMAD router ratchet |
Tightened bmad into a packet-first BMAD/BMM front door. It now starts from the packet already in hand (idea notes, PRD, architecture draft, review feedback, brownfield repo state, or milestone pressure), chooses one next artifact or gate, keeps review/runtime boundaries explicit, adds references/intake-packets-and-route-outs.md, expands evals with a brownfield mixed-state case, and refreshes discovery wording so bmad stops reading like a generic phase catalog. |
| bmad-gds: game producer/orchestration rewrite |
Reframed bmad-gds from a generic phase catalog into a practical game-production coordination skill. It now acts as the repo's game-cluster orchestrator: normalizes mixed packets (idea, GDD, playtest notes, bug/build issues, launch targets), chooses one operating mode, produces one milestone-aware coordination brief, and routes detailed work to game-demo-feedback-triage, game-build-log-triage, game-performance-profiler, steam-store-launch-ops, task-planning, or bmad-idea as needed. Added references/operating-modes.md, references/scope-boundaries.md, and evals/evals.json without increasing the skill count. |
🆕 What's New in v2026-04-08
| Change |
Details |
| graphify: repo and corpus knowledge-graph skill |
Added a dedicated graphify skill for turning repositories or mixed corpora into persistent knowledge-graph artifacts with GRAPH_REPORT.md, graph.json, and HTML visualization. Covers the tested Python API pipeline, graph queries, graph-backed architecture discovery, and assistant install flows; includes references/overview.md and evals/evals.json. 84 → 85 skills. |
| llm-wiki: persistent LLM-maintained markdown wiki skill |
Added a dedicated llm-wiki skill for turning raw sources into a compounding Obsidian or markdown knowledge base. It bootstraps a vault with raw/, wiki/, index.md, log.md, and AGENTS.md; ships helper scripts for bootstrap, Scrapling-powered URL ingest, query filing, and linting; and pushes schema, ingest, filing, and scaling detail into focused references. Includes evals/ plus skill-autoresearch-llm-wiki/ baseline, changelog, results, and dashboard artifacts. 82 → 83 skills. |
| rtk: Rust Token Killer setup and operations skill |
Added a dedicated rtk skill for installing, verifying, and initializing Rust Token Killer across Claude Code, Codex, Gemini CLI, Cursor, Copilot, Windsurf, Cline, and OpenCode. The skill starts with rtk gain verification, handles the common wrong-package collision, ships install/init/status wrapper scripts, and pushes deeper platform and troubleshooting details into focused reference docs. Includes evals/ plus skill-autoresearch-rtk/ baseline, changelog, results, and dashboard artifacts. 81 → 82 skills. |
🆕 What's New in v2026-03-30
| Change |
Details |
| harness: agent team & skill architect meta-skill |
Added a dedicated harness skill for designing domain-specific agent teams and generating the skills they use. Covers domain analysis, architecture pattern selection (pipeline, fan-out/fan-in, expert pool, producer-reviewer, supervisor, hierarchical delegation), .claude/agents/ and .claude/skills/ file generation, orchestration workflow definition, and validation with trigger evals and dry-run testing. Includes install.sh, validate-harness.sh scripts, and 5 reference docs. 80 → 81 skills. |
🆕 What's New in v2026-03-28
| Change |
Details |
| obsidian-cli: routing-first Obsidian desktop automation |
Hardened obsidian-cli into a routing-first front door for official Obsidian desktop automation: choose CLI command/TUI mode, developer-command mode, or official obsidian:// URI handoff first; prefer deterministic vault= + path= targeting; and route headless sync, raw filesystem writes, or richer plugin/API automation away instead of overclaiming CLI coverage. Refreshed install/troubleshooting guidance plus new intake/route-out reference included. |
| scrapling: routing-first adaptive web scraping skill |
Added and then hardened the dedicated scrapling skill so it now routes users into the lightest workable mode first: parser-only HTML, HTTP fetch, JS-rendered browser fetch, protected-target stealth, CLI/MCP operator flows, or full spiders. The implementation includes install/extract/MCP wrapper scripts plus focused references for fetchers, parser behavior, CLI/MCP, spiders, and intake-packet route-outs. 78 → 79 skills. |
| strix: AI-driven application security testing skill |
Added a dedicated strix skill for operating the Strix CLI end-to-end: install and Docker preflight, STRIX_LLM provider setup, local/GitHub/live target scans, quick/standard/deep mode selection, headless CI/CD usage, and clear separation between this repo's skill and Strix internal security skills. 77 → 78 skills. |
🆕 What's New in v2026-03-22
| Change |
Details |
| bmad-orchestrator renamed to bmad |
bmad-orchestrator skill folder renamed to bmad. Simplified to core BMAD workflow orchestration (Analysis → Planning → Solutioning → Implementation). Use keyword bmad as before. |
| Removed copilot-coding-agent |
copilot-coding-agent skill removed. 77 skills total. |
🆕 What's New in v2026-03-19
| Change |
Details |
| clawteam: ClawTeam runtime operator router |
Tightened clawteam into a packet-first ClawTeam runtime skill: choose one honest operator packet (manual-team, template-launch, monitor-recover, or profile-setup) before commands, keep tmux/worktree runtime reality explicit, and route generic orchestration or board-governance requests outward. |
| obsidian-plugin: Obsidian plugin development skill |
Build, validate, and publish Obsidian plugins. Covers all 27 eslint-plugin-obsidianmd rules, interactive boilerplate generator (create-plugin.js), memory management, type safety, accessibility (MANDATORY), CSS variables, vault API, and community submission validation. 75 → 76 skills. |
jeo v1.6.0: .jeo planning ledger flow |
JEO now creates a project-local .jeo/ folder and uses it as a durable planning/development/QA ledger: long-term.md, short-term.md, planned.md, progress.md, history.md, plus queued/active task files. Completed task files are summarized into history then removed; follow-up work can be queued without resetting the workflow. |
| skill-autoresearch: eval-driven skill optimization |
New skill for improving an existing SKILL.md with binary evals, mutation loops, baseline scoring, and dashboard/changelog artifacts. Keeps the original autoresearch ML workflow separate. 76 → 77 skills. |
| firebase-cli: Firebase platform operator hardening |
Upgraded firebase-cli into a routing-first Firebase operator anchor for install/auth, bootstrap/config, Emulator Suite workflows, scoped deploy/release flows, and admin/data operations. Added focused references for routing, bootstrap/auth, emulators/release, and admin tasks; refreshed evals/compact wording; and corrected the npm-path Node.js baseline to current firebase-tools requirements. |
| google-workspace, langsmith, react-grab added |
3 new skills: Google Workspace REST API automation, LangSmith LLM observability/evaluation, react-grab React element context capture. 71 → 74 skills. |
| research-paper-writing: ML/CV/NLP paper writing skill |
Academic paper and rebuttal workflow for Abstract, Introduction, Method, Experiments, figures/tables, reviewer responses, and camera-ready revision. Claim-evidence alignment, section planning, and reviewer-risk checks. From Prof. Peng Sida's notes plus repo support hardening. 70 → 71 skills. |
| Removed ai-tool-compliance and llm-monitoring-dashboard |
Removed ai-tool-compliance (internal compliance automation) and llm-monitoring-dashboard. 72 → 70 skills. |
| Removed deprecated agent-development skills |
Removed agent-configuration, agent-evaluation, agentic-development-principles, agentic-principles, agentic-workflow. 80 → 72 skills. |
| Removed deprecated image/media skills |
Removed image-generation, image-generation-mcp, pollinations-ai. Use video-production as the canonical programmable-video skill; remotion-video-production remains as the compatibility alias for explicit Remotion naming. |
| autoresearch: Karpathy autonomous ML experiment skill |
Human-written program.md, agent-edited train.py, fixed 5-minute GPU runs, and val_bpb keep/revert ratcheting for real ML search. Now explicitly routes prompt / app eval work away to skill-autoresearch or eval platforms, and includes scripts/, references/, and evals/. |
| jeo v1.2.3: plannotator-plan-loop.sh all-platform hardening |
Cross-platform temp dir fallback, dedicated port PLANNOTATOR_PORT=47291, probe_plannotator_port() + wait_for_listen(), browser-crash retry up to 3 times, structured jeo-blocked.json output. |
| survey: artifact-validator hardening |
Tightened survey into a smaller artifact-contract-first research anchor, moved verbose output templates into a dedicated reference, and added scripts/validate_survey_artifacts.py so .survey/{slug}/ folders can be checked mechanically before planning or implementation. Platform topics still normalize as settings/rules/hooks, with hooks treated as optional wrappers around the same portable validator. |
| presentation-builder: packet-first deck handoff hardening |
Tightened presentation-builder into a smaller routing-first deck artifact anchor. It now chooses one deck mode, one smallest useful artifact packet (outline-brief, storyboard, review-ready-html, export-handoff, or sync-packet), and one honest last-mile surface (HTML viewer, PPTX, PDF, Google Slides, or Figma Slides); adds references/artifact-packets-and-last-mile-handoffs.md; refreshes eval coverage; and syncs compact/discovery surfaces so the skill matches real deck workflows instead of acting like a giant slides essay. |
📦 Installation
Step 0: Install skills CLI
npm install -g skills
skills --version
For LLM Agents
curl -s https://raw.githubusercontent.com/akillness/oh-my-skills/main/setup-all-skills-prompt.md
Claude Code
npx skills add https://github.com/akillness/oh-my-skills \
--skill jeo --skill omc --skill plannotator --skill agentation \
--skill ooo --skill vibe-kanban
Gemini CLI
npx skills add https://github.com/akillness/oh-my-skills \
--skill jeo --skill ohmg --skill ooo --skill vibe-kanban
gemini extensions install https://github.com/akillness/oh-my-skills
Codex CLI
npx skills add https://github.com/akillness/oh-my-skills \
--skill jeo --skill omx --skill ooo
# Claude Code — jeo hook setup
bash ~/.agent-skills/jeo/scripts/setup-claude.sh
# Gemini CLI — jeo hook setup
bash ~/.agent-skills/jeo/scripts/setup-gemini.sh
# oh-my-claudecode
/plugin marketplace add https://github.com/Yeachan-Heo/oh-my-claudecode
/plugin install oh-my-claudecode
setup omc
📚 Skills List
Full manifest: .agent-skills/skills.json · each folder's SKILL.md · 90 local skill folders = 90 total installable skills
🎯 Core Orchestration (11)
| Skill |
Keyword |
Platform |
Description |
jeo |
jeo, annotate |
All |
Packet-first orchestration front door with .jeo ledger: plan gate → runtime handoff → verify → cleanup |
omc |
omc, autopilot, ralph, ulw, ccg, /team, omc team, omc ask, cancelomc |
Claude |
Claude-first orchestration router for oh-my-claudecode — identifies marketplace plugin vs shell CLI vs local --plugin-dir topology first, distinguishes plugin slash skills from the omc shell CLI, handles duplicate-install/recovery/state issues, and routes adjacent work to jeo, ralphmode, omx, ohmg, and browser-review skills |
harness |
harness, build a harness |
All |
Meta-skill: design domain-specific agent teams, generate .claude/agents/ + .claude/skills/ files, validate harness |
omx |
omx, $plan, $ralph, $team, $deep-interview, $ralplan |
Codex |
Multi-agent workflow layer for Codex CLI (v0.11.10) — 30+ agents, 35+ skills, tmux team runtime, omx explore/sparkshell |
ohmg |
ohmg, oh-my-agent, oma, .agents |
Gemini |
Gemini / Antigravity entry for the portable oh-my-agent harness (.agents source of truth, native Gemini projection, cross-vendor-ready layout) |
ooo |
ooo, ouroboros, ooo ralph |
All |
Ouroboros spec-first development loop — Socratic interview, immutable seed/spec, drift-aware execution, persistent completion until verification passes. Plugin: claude plugin marketplace add Q00/ouroboros |
bmad |
bmad, workflow-init, workflow-status |
All |
Packet-first BMAD/BMM front door — classify the current packet, choose the next artifact or gate, and route runtime / review / execution detail outward |
bmad-gds |
bmad-gds |
All |
Game-production orchestrator — turn ideas, GDDs, playtest notes, bugs, and launch beats into one milestone-aware next artifact |
bmad-idea |
bmad-idea |
All |
Pre-planning idea router — turn rough product, GTM, consulting, or game ideas into one concept artifact and the next handoff |
survey |
survey |
All |
Bounded pre-implementation landscape scan with reusable .survey/{slug}/ artifacts plus validator-backed artifact-contract checks |
📋 Planning & Review (5)
| Skill |
Keyword |
Description |
plannotator |
plan |
Visual approval gate for agent plans/diffs — annotate, approve, request changes, or save reviewed plans |
agentation |
annotate |
Exact rendered-UI feedback router — choose copy-paste review, watch-loop sync, self-driving critique, or platform setup |
agent-browser |
agent-browser |
Fresh-session browser verification anchor — clean disposable browser, snapshot refs, and explicit before/after evidence |
playwriter |
playwriter |
Running-browser automation for authenticated Chrome sessions and MCP browser reuse |
vibe-kanban |
kanbanview |
Coding-board control plane for bounded coding cards, tracker-linked workspaces, review queues, worktree isolation, and PR handoff |
🤖 Agent Development (2)
| Skill |
Description |
Platforms |
prompt-repetition |
Decision-first prompt repetition skill for non-reasoning/lightweight LLMs — long-context retrieval, options-first MCQ, position-sensitive lookup, and explicit route-outs to retrieval or stronger models |
All |
skill-standardization |
Validate/rewrite SKILL.md, canonicalize duplicates, and keep repo-root validator flows plus derived discovery surfaces (skills.json, README/setup, SKILL.toon) in sync |
All |
⚙️ Backend (5)
| Skill |
Description |
Platforms |
api-design |
Contract-first REST/GraphQL API design, compatibility review, and handoff |
All |
api-documentation |
Developer-facing API docs anchor for reference portals, quickstarts, SDK/webhook guides, truthful examples, and auth/error guidance |
All |
authentication-setup |
Product-auth setup routing across hosted/framework-native/platform-native auth, sessions/JWTs, org data, and enterprise SSO handoff |
All |
backend-testing |
Packet-first backend testing for coverage plans, fixture/reset strategy, contract/API protection, flaky-suite stabilization, and local-vs-CI lane splits |
All |
database-schema-design |
Packet-first storage-model and migration-safety design for relational/document/hybrid schemas, queryable-vs-flexible fields, and route-outs to API/auth/testing/reporting neighbors |
All |
🎨 Frontend (11)
| Skill |
Description |
Platforms |
design-system |
Canonical frontend UI-system anchor for token governance, visual-language rules, primitive naming, and cross-surface system direction; routes component API, responsive layout, accessibility remediation, and broad UI critique to adjacent skills |
All |
frontend-design-system |
Compatibility alias for design-system when legacy tooling or exact-name workflows still expect the old name |
All |
google-design |
DESIGN.md format for AI coding agents — machine-readable YAML design tokens + markdown rationale; lint (WCAG AA), diff versions, export to Tailwind/DTCG |
All |
stitch-skills |
Agent Skills for Stitch MCP — generate high-fidelity UI screens, multi-page websites, DESIGN.md docs, enhance prompts, convert to React/shadcn-ui, Remotion walkthrough videos. Plugin: claude plugin marketplace add google-labs-code/stitch-skills |
All |
react-best-practices |
Measurement-led React & Next.js performance audits for waterfalls, bundle size, RSC/client boundaries, hydration, rerender churn, and slow routes |
All |
react-grab |
Browser element context capture — point at UI element, copy React component name, file path, HTML to clipboard for AI agents |
All |
vercel-react-best-practices |
Compatibility alias for react-best-practices when legacy tooling or exact-name workflows still expect the Vercel variant |
Claude · Gemini · Codex |
responsive-design |
Routing-first responsive layout strategy for page-shell, component-slot, dense-data, media, and reflow-verification packets |
All |
state-management |
React/fullstack ownership-packet decisions across local, Context, URL/form, client-store, and server-state/router data layers |
All |
ui-component-patterns |
Routing-first reusable-component architecture for primitive-boundary, slot-anatomy, controlled-ownership, alternate-root, and docs-verification packets |
All |
web-accessibility |
Routing-first accessibility remediation and verification for semantics, keyboard/focus, labels/announcements, reflow, media alternatives, and routed-app feedback |
All |
web-design-guidelines |
Broad web UI audit for hierarchy, clarity, consistency, states, responsiveness basics, and accessibility basics |
All |
🔍 Code Quality (5)
| Skill |
Description |
Platforms |
code-refactoring |
Behavior-preserving structural cleanup, decomposition, duplication removal, and codemod planning |
All |
code-review |
Evidence-first diff / PR review with severity, missing-proof checks, and route-outs |
All |
debugging |
Routing-first diagnosis for concrete bugs, regressions, flaky failures, and env-specific behavior; routes raw logs to log-analysis and perf-only work to performance-optimization |
All |
performance-optimization |
Artifact-first measurement-led bottleneck analysis and tuning across latency, throughput, memory, bundle, CWV, and frame-budget work |
All |
testing-strategies |
Packet-first validation policy for merge-gate truth, release-only proof, scheduled breadth, and cross-domain test-policy handoffs |
All |
🏗 Infrastructure (13)
| Skill |
Description |
Platforms |
deployment-automation |
Release-execution anchor for preview releases, staging/prod promotion, rollout strategy, post-deploy verification, rollback response, and release hardening; routes CI authoring to workflow-automation, machine setup to system-environment-setup, and Vercel-specific operations to vercel-deploy |
All |
environment-setup |
App-config compatibility skill for .env layout, env precedence, validation, and secret handoff; routes broader runnable-machine setup to system-environment-setup |
All |
firebase-ai-logic |
Direct Firebase app/client SDK lane for Gemini-powered features, streaming, structured output, and App Check-aware in-app integration; routes backend orchestration to genkit |
Claude · Gemini |
firebase-cli |
Firebase platform/operator anchor for install/auth, bootstrap/config, Emulator Suite workflows, scoped deploy/release, App Hosting, and admin/data ops; routes backend AI workflow orchestration to genkit and direct app SDK integration to firebase-ai-logic |
All |
genkit |
Packet-first backend AI workflow anchor for deciding whether a feature needs a reusable server-owned flow, Genkit eval/tracing, or a fallback to plain SDK routes / survey; routes direct app SDK work to firebase-ai-logic and Firebase operator tasks to firebase-cli |
Claude · Gemini |
looker-studio-bigquery |
Packet-first BigQuery dashboard/reporting lane for dashboard-spec, slow-dashboard, refresh-shape, audience-split, and exec-handoff; routes KPI interpretation to data-analysis |
All |
monitoring-observability |
Packet-first telemetry design/review for service health, telemetry rollout, alert/dashboard audits, pipeline trust, and live-ops visibility |
All |
scrapling |
Routing-first adaptive web scraping: choose parser-only, HTTP fetch, JS browser, stealth escalation, MCP, or spiders from one intake packet |
All |
rtk |
Rust Token Killer installation and agent setup - rtk gain verification, package-collision repair, agent-specific rtk init, and direct compact shell wrappers |
All |
security-best-practices |
Routing-first web/application/API hardening that classifies the missing security layer (browser policy, cookies/CSRF, abuse, validation, secrets, verification) before recommending one bounded hardening brief |
All |
strix |
Strix CLI for AI-driven application security testing - Docker preflight, LLM provider setup, local/GitHub/live target scans, scan modes, and CI/CD usage |
All |
system-environment-setup |
Canonical broader environment-setup skill for runnable repos, toolchains, Docker/devcontainers, local services, onboarding, and setup drift diagnosis |
All |
vercel-deploy |
Vercel-specific operator skill for linked-project preview/prod deploys, staged promote flows, aliases/domains, env-scope fixes, and rollback response |
All |
📝 Documentation (5)
| Skill |
Description |
Platforms |
changelog-maintenance |
Routing-first release-history anchor for changelogs, release notes, migration updates, and lightweight patch-note packets |
All |
presentation-builder |
Packet-first deck artifact anchor for investor / roadmap / launch / architecture-demo / workshop / game-pitch decks, with honest last-mile handoff to HTML review, PPTX, PDF, Google Slides, or Figma Slides |
All |
research-paper-writing |
ML/CV/NLP academic paper + rebuttal workflow — abstract/introduction/method/experiments, figure-table support, claim-evidence alignment, reviewer response, camera-ready revision |
All |
technical-writing |
Internal technical docs anchor for specs, architecture docs, ADRs, runbooks, migration guides, and developer-facing implementation notes |
All |
user-guide-writing |
Mode-selecting user-docs anchor for onboarding guides, tutorials, task how-to articles, FAQs, help-center updates, and release-facing help refresh packets |
All |
📊 Project Management (4)
| Skill |
Description |
Platforms |
sprint-retrospective |
Routing-first retrospective anchor for sprint retros, milestone postmortems, remote/hybrid facilitation, and dead-action-item recovery |
All |
standup-meeting |
Routing-first coordination-cadence anchor for deciding whether daily, async, hybrid, lighter, or no recurring standup is justified before choosing a standup mode |
All |
task-estimation |
Routing-first estimate packet anchor for story points, t-shirt sizing, split/spike guidance, and forecast-safe uncertainty framing across software, GTM, and game work |
All |
task-planning |
Packet-first planning anchor for backlog cleanup, feature slicing, sprint/milestone prep, and release packets with explicit route-outs to estimation, boards, review, and pre-planning framing |
All |
🔭 Search & Analysis (7)
| Skill |
Description |
Platforms |
autoresearch |
Karpathy autonomous ML search front door — choose setup / program.md / bounded loop / results interpretation / constrained-hardware mode, preserve immutable prepare.py + 300s + val_bpb, route prompt/skill eval elsewhere |
All |
skill-autoresearch |
Repo-local skill ratcheting loop: choose one packet (ratchet eligibility, readiness, charter, baseline, mutation, support-sync, final report), allow no ratchet justified, freeze evals, keep or revert by score, and route hosted eval / ML autoresearch work outward |
All |
codebase-search |
Routing-first repo navigation: choose one search packet for definitions/references, config/content ownership, entry-point discovery, or impact mapping before debugging/refactoring |
All |
data-analysis |
Decision-first dataset analysis for exports, experiments, telemetry, and KPI explanation |
All |
langsmith |
Routing-first LangSmith skill: choose one packet for trace-debug, evals, review queues, prompt-registry decisions, or cross-service propagation before touching SDK code |
All |
log-analysis |
Routing-first log triage: choose one evidence packet for app, container/pod, browser+API, CI cascade, JSON/event, or security-signal logs before debugging/observability work |
All |
pattern-detection |
Routing-first pattern/anomaly hunting: choose text-prefilter, structural-code-rule, log-event-pattern, or metric-anomaly before deeper analysis |
All |
| Skill |
Description |
Platforms |
remotion-video-production |
Compatibility alias for video-production when legacy tooling or explicit Remotion naming still expects the old skill |
All |
video-production |
Canonical programmable-video / automated-video production skill for Remotion, template APIs, content repurposing, and QA handoffs |
All |
📢 Marketing (2)
| Skill |
Description |
Platforms |
marketing-automation |
Canonical broad marketing front door — choose one operating mode, one primary lane, and one reusable operator packet with owner, dependencies/approvals, and proof across launch, conversion, lifecycle, acquisition/content, and measurement work |
All |
marketing-skills-collection |
Compatibility alias for marketing-automation in legacy prompt packs and catalogs |
All |
🎮 Game Development (5)
| Skill |
Description |
Platforms |
game-build-log-triage |
Unity/Unreal build, cook, package, editor, signing, and CI log triage — isolate the first actionable engine/build failure |
All |
game-ci-cd-pipeline |
Game pipeline packet router — classify branch-gate vs nightly/package-candidate vs release/certification lane, then choose setup, stage split, cache policy, preflight, artifact/release hygiene, or CI-signal hardening |
All |
game-demo-feedback-triage |
Turn playtest/demo/community feedback into weighted themes, fix-first priorities, and explicit handoffs |
All |
game-performance-profiler |
Unity/Unreal frame-time triage — bottleneck-first profiling brief, quick packets, benchmark routes, target-device review, and deliberate profiler escalation |
All |
steam-store-launch-ops |
Packet-first Steam launch router — choose page-promise audit, wishlist signal check, demo readiness, event timing workback, or launch-ops runbook |
All |
🔧 Utilities (11)
| Skill |
Description |
Platforms |
fabric |
Routing-first Fabric operator skill for stdin/files/transcripts/logs → reusable patterns, custom pattern packs, shell pipelines, and fabric --serve workflows |
All |
file-organization |
Decision-first repo structure guidance — choose framework defaults vs feature folders, monorepo boundaries, docs taxonomy, or engine-aware game layouts |
All |
git-submodule |
Routing-first Git submodule workflow — choose submodule vs subtree/vendor/package delivery, bootstrap recursively, sync or advance pointers safely, avoid detached-HEAD traps, and handle CI / hosted-platform checkout limits |
All |
git-workflow |
Routing-first local Git workflow — branch hygiene, selective staging, commit cleanup, merge-vs-rebase choice, conflict resolution, lease-safe pushes, and recovery |
All |
google-workspace |
Decision-first Google Workspace operations — choose Apps Script vs REST API vs Admin SDK, auth model, and cross-service flows across Docs, Sheets, Slides, Drive, Gmail, Calendar, Chat, and Forms |
All |
graphify |
Routing-first durable graph skill — choose assistant-native install, local build, refresh, graph query, or structural fallback for repo/corpus graphs and persistent GRAPH_REPORT.md / graph.json outputs |
All |
llm-wiki |
Persistent markdown wiki maintenance for Obsidian or git-tracked vaults — raw sources, source summaries, query filing, lint passes, and optional Scrapling/qmd helpers |
All |
npm-git-install |
Routing-first Node package delivery skill for npm / pnpm / Yarn / Bun — choose temporary Git bridge, SHA pin, tarball, workspace, or publish-first handoff safely |
All |
obsidian-cli |
Route Obsidian desktop automation — official CLI command/TUI mode, developer commands, official obsidian:// handoff, and explicit route-outs for headless or plugin/API-heavy jobs |
All |
obsidian-plugin |
Obsidian plugin development — 27 ESLint rules, boilerplate generator, accessibility, submission validation |
All |
opencontext |
Routing-first active project/repo memory — choose memory-layer choice, load-context, search-context, store-conclusions, setup-integration, or repo-packer route-out for manifests, stable links, cross-agent handoff packets, and highest-confidence-source / freshness checks when notes overlap |
All |
workflow-automation |
Routing-first repo workflow automation — choose one mode for task-entrypoints, bootstrap/onboarding, local-CI parity, hook guardrails, maintenance bots, or workflow cleanup without drifting into environment/deploy work |
All |
TOON (Token-Oriented Object Notation) compresses the skill catalog and auto-injects it into every prompt. 40-50% token savings vs JSON/Markdown.
| Platform |
File |
Mechanism |
| Claude Code |
~/.claude/hooks/toon-inject.mjs |
UserPromptSubmit hook — 26-37ms |
| Gemini CLI |
~/.gemini/hooks/toon-skill-inject.sh |
includeDirectories session load |
| Codex CLI |
~/.codex/skills-toon-catalog.toon |
Static catalog |
- Tier 1 (always): Skill catalog index (~875-3,500 tokens) — names + descriptions + tags
- Tier 2 (on-demand): Individual SKILL.toon content (~292 tokens/skill, max 3)
jeo — Integrated Agent Orchestration
Keyword: jeo · annotate | Platforms: Claude · Codex · Gemini · OpenCode
Packet-first orchestration front door: choose the right JEO packet, preserve .jeo ledger truth, and hand work to the correct owner instead of expanding every runtime inline.
JEO keeps the shared contract — plan gate, runtime handoff, verification requirements, submit-gated UI review, cleanup, and resumable .jeo / machine state — while routing specialist work to sibling skills.
| Packet / Phase |
Owner |
Description |
| Bootstrap / Resume |
JEO scripts + .jeo/ |
Initialize or recover durable ledger and machine state |
| Plan / Planning |
ooo + plannotator |
Shape the plan and get approval without reopening unchanged work |
| Runtime handoff / Execute |
omc / omx / ohmg / truthful bmad fallback |
Keep runtime-native config and execution in the runtime skill |
| Verify / QA |
agent-browser |
Record browser / QA evidence before claiming completion |
| Verify UI / annotate |
agentation |
Wait for explicit submit, then process UI feedback |
| Cleanup |
JEO scripts + worktree-cleanup.sh |
Summarize, queue follow-up work, and clean worktrees |
plannotator — Visual Plan Review
Keyword: plan | Docs | GitHub
Browser UI for annotating AI plans. Approve or send structured feedback in one click. Works with Claude Code, OpenCode, Gemini CLI, and Codex CLI.
bash scripts/install.sh --all
ooo — Ouroboros Specification-First Development
Keyword: ooo, ouroboros, ooo ralph | Docs | GitHub
Socratic interview → immutable seed/spec → execute against the contract → verify before done → keep looping until completion is actually verified. Installable as a Claude Code plugin or via pip.
# Plugin install (Claude Code)
claude plugin marketplace add Q00/ouroboros
# pip
pip install ouroboros-ai[all]
# Skill install (any platform)
npx skills add https://github.com/akillness/oh-my-skills --skill ooo
# Usage
ouroboros init start "I want to build a task management CLI"
ouroboros run workflow seed.yaml
ouroboros run resume
ouroboros tui monitor
google-design — DESIGN.md for AI Coding Agents
Keyword: google-design, DESIGN.md | Docs | GitHub
Machine-readable design system format (YAML tokens + markdown rationale) that gives AI coding agents a persistent, structured understanding of a visual identity. Lint for WCAG compliance, diff versions, export to Tailwind or W3C DTCG format.
# npm install
npm install @google/design.md
# Skill install (any platform)
npx skills add https://github.com/akillness/oh-my-skills --skill google-design
# Usage
npx @google/design.md spec > DESIGN.md # bootstrap
npx @google/design.md lint DESIGN.md # validate + WCAG check
npx @google/design.md export --format tailwind DESIGN.md > tailwind.theme.json
npx @google/design.md diff old.md new.md # track changes
stitch-skills — Agent Skills for Stitch MCP
Keyword: stitch, stitch-design, stitch-loop, enhance-prompt | Docs | GitHub
AI-powered UI design generation, prompt refinement, and screen-to-code workflows via the Stitch MCP server. Generate high-fidelity screens, multi-page websites, DESIGN.md docs, React/shadcn-ui components, and Remotion walkthrough videos.
# Plugin install (Claude Code)
claude plugin marketplace add google-labs-code/stitch-skills
# Skill install (any platform)
npx skills add google-labs-code/stitch-skills --skill stitch-design --global
npx skills add google-labs-code/stitch-skills --skill enhance-prompt --global
# Install from oh-my-skills
npx skills add https://github.com/akillness/oh-my-skills --skill stitch-skills
vibe-kanban — AI Agent Kanban Board
Keyword: kanbanview | Docs | GitHub
Coding-board control plane for bounded coding cards: keep GitHub Projects / Linear / Jira as the PM source of truth when needed, run isolated workspaces or worktrees for actual coding execution, keep human review explicit, and hand off cleanly to PRs.
npx vibe-kanban
🌐 Recommended Harness OSS
| Repository |
Stars |
Description |
| AutoGPT |
182k |
Accessible AI platform for continuous agents |
| AutoGen |
55.4k |
Microsoft multi-agent conversation framework |
| CrewAI |
45.7k |
Role-playing autonomous AI agent orchestration |
| smolagents |
25.9k |
HuggingFace code-thinking agent library |
| agency-agents |
21.2k |
61 specialized AI agents across 9 divisions |
| revfactory/harness |
meta-skill |
Agent team & skill architect plugin / scaffold |
Install & integration notes → docs/harness/README.md · packaged skill → .agent-skills/harness/SKILL.md
📁 Structure
.
├── .agent-skills/ ← 90 skill folders (each with SKILL.md + SKILL.toon)
├── docs/ ← detailed guides (bmad, omc, plannotator, ooo, ...)
├── install.sh
├── setup-all-skills-prompt.md
├── README.md ← English (this file)
└── README.ko.md ← 한국어
📎 References

⭐ If this helps you, star the repository!
gh api --silent --method PUT /user/starred/akillness/oh-my-skills >/dev/null 2>&1 || true