fth-skills

🛠️ fth-skills

Curated AI agent skills for coding workflows, decision-making, and agentic task safety

Reusable instruction packs for the skills.sh ecosystem — works with Claude Code, Cursor, Copilot, Gemini CLI, and more.

⚡ Quick Install

skills.sh

# Install all skills
npx skills add fatih-developer/fth-skills

# Install a single skill
npx skills add fatih-developer/fth-skills --skill <skill-name>

Hermes Agent

# Add this repository as a reusable skill tap
hermes skills tap add fatih-developer/fth-skills

# Search the tap from Hermes
hermes skills search <query>

# Install a single skill from the tap
hermes skills install fatih-developer/fth-skills/<skill-name>

# Example
hermes skills install fatih-developer/fth-skills/project-analyzer

ClawHub

# Install and authenticate the ClawHub CLI
npm i -g clawhub
clawhub login
clawhub whoami

# Validate the public ClawHub publish list
python scripts/publish_clawhub.py --check-only

# Dry-run all public skills listed in skills.sh.json
python scripts/publish_clawhub.py

# Dry-run one skill
python scripts/publish_clawhub.py --skill project-analyzer

# Publish one skill after reviewing the dry-run output
python scripts/publish_clawhub.py --skill project-analyzer --execute

# Publish all public skills after a successful single-skill smoke test
python scripts/publish_clawhub.py --execute

# Inspect scanner results and published files
npx clawhub inspect <skill-slug> --files

🎯 Available Skills (79 Public)

Skills are grouped into ecosystems. Each ecosystem has a hub skill (ecosystem-*) that picks a workflow, checks which member skills are installed, runs them in order with explicit handoffs, and falls back gracefully when a member is missing. Install a hub to get the routing, then add members as you need them.

Agent Orchestration (16 + hub)

Plan, execute, recover, remember, and review multi-step agent work; connect external agents safely. Workflows: orch-complex-task, orch-code-change, orch-resilience, orch-agent-connectivity, orch-knowledge.

Reasoning, Planning Gates & Prompting (10 + hub)

Clarify intent, stress-test plans, compare perspectives, and craft prompts, without stacking redundant question gates. Workflows: reason-clarify, reason-decision, reason-plan-review, reason-prompting.

API Domain (10 + hub)

Design, secure, scale, and evolve REST, GraphQL, gRPC, and event-driven APIs with a contract-first workflow. Workflows: api-design, api-security-scale, api-evolution, api-production-readiness.

Database Domain (12 + hub)

Model, secure, optimize, migrate, and operate SQL databases (PostgreSQL, MySQL, SQLite) with data integrity first. Workflows: db-new-schema, db-performance, db-evolution, db-compliance, db-scaling.

Mobile App Domain (10 + hub)

Architect, audit, release, and monitor iOS, Android, React Native, and Flutter apps. Workflows: mobile-pre-release-audit, mobile-release, mobile-feature-architecture, mobile-post-launch.

Web, Design & Delivery (7 + hub)

Design product-grade web interfaces, keep design systems consistent, make sites visible to search and AI answer engines, and deploy them. Workflows: web-foundations, web-build-page, web-visibility, web-review, web-de-slop, web-deploy.

Product & Growth (3 + hub)

Take a product from focus to market: B2B go-to-market, naming, creative content, and discoverability. Workflows: product-launch, product-naming, product-content.

Security (3 + hub)

Audit skills before deployment, guard agent runtimes against injection and exfiltration, and learn from incidents. Workflows: sec-skill-audit, sec-runtime-guard, sec-full-lifecycle.

Deprecated aliases

Kept so existing installs keep working. They route to the canonical skill and never trigger implicitly.

🧩 How the Ecosystems Work

graph LR
    U["User request"] --> H["ecosystem-orchestration<br/>(hub + index)"]
    H --> E["ecosystem-&lt;domain&gt;<br/>router"]
    E --> P["Preflight:<br/>installed or missing?"]
    P -->|installed| S["Member skill"]
    P -->|missing| F["Fallback: run the step<br/>from the workflow map"]
    S --> A["Handoff artifact<br/>(contract v1)"]
    F --> A
    A --> N["Next step"]
    N -->|irreversible| G["checkpoint-guardian"]
    N -->|failed| R["error-recovery"]

1. One source of truth. Every ecosystem keeps its map in skills/ecosystem-<domain>/references/workflows.json: members, workflows, steps (with Does, Done when, conditions, parallel groups), direct handoffs, and — for reasoning — gate precedence.
2. Generated, never drifting. scripts/build_ecosystems.py renders the router, catalog, and workflow sections of each hub, the 🔗 Next Steps & Handoffs section of every member skill, the ecosystem index inside task-decomposer and parallel-planner, the skills.sh.json groupings, and this README catalog. --check fails CI when anything is out of date.
3. Graceful degradation. A missing member never blocks a workflow. The hub offers to install it (scripts/install_all.py is a dry run until --execute) or runs the step itself in fallback mode.
4. Handoff contract. Steps pass results through a small JSON envelope (skill, workflow, created_at, inputs, summary, outputs, next), in files for workflow runs and inline for quick questions. See references/handoff-contract.md in any hub.
5. Safety built in. Irreversible or production-impacting steps stop at checkpoint-guardian; failures go to error-recovery; only one clarification gate runs per request.

📁 Repository Structure

fth-skills/
├── skills/
│   ├── ecosystem-<domain>/            # hub: router + workflow map
│   │   ├── SKILL.md                   # generated blocks + hand-written intro
│   │   ├── references/
│   │   │   ├── workflows.json         # source of truth for the domain
│   │   │   └── handoff-contract.md    # generated from scripts/templates
│   │   └── scripts/install_all.py     # generated; dry-run installer
│   └── <skill-name>/
│       ├── SKILL.md                   # frontmatter + instructions + generated handoffs
│       ├── agents/openai.yaml         # required for skills.sh discovery
│       ├── references/ templates/ scripts/ assets/   # optional
│       └── evals/                     # optional trigger and behavior evals
├── scripts/
│   ├── build_ecosystems.py            # renders every ecosystem-derived artifact
│   ├── validate_curated_skills.py     # structural validator (run in CI)
│   ├── templates/ecosystem/           # installer + handoff contract templates
│   └── ...
├── skills.sh.json                     # generated groupings
└── .github/workflows/validate.yml     # CI: build --check + validator + tests

🤝 Contributing

1. Create skills/<skill-name>/SKILL.md with frontmatter. The description must say what the skill does and when to use it:

   ---
   name: skill-name
   description: "What it does. Use when ..."
   ---
   

2. Add skills/<skill-name>/agents/openai.yaml with display_name, a complete short_description (no trailing ...), and a default_prompt that mentions $skill-name.
3. Add the skill to one ecosystem's references/workflows.json (as a member, and in workflows or handoffs where it belongs).
4. Regenerate and validate:

   python scripts/build_ecosystems.py
   python scripts/validate_curated_skills.py
   

5. Commit the regenerated files together with your change and open a PR. CI runs the same checks.

License

MIT