Skip to content

About

Agent Skills for serious software engineering: commits, code review, debugging, security audits, Docker & CI.

Topics

Resources

Code of conduct

Contributing

Security policy

Stars

1 star

Watchers

0 watching

Forks

Repository files navigation

⚒️ SkillForge

Agent Skills for serious software engineering.
18 tested skills that turn your AI coding agent into a disciplined senior engineer:
clean commits, real code reviews, root-cause debugging, security audits, production Docker and CI.

Validate Skills License: MIT PRs welcome GitHub stars

English · فارسی


Why SkillForge?

AI coding agents are fast, but left alone they guess. They write vague commit messages, "fix" symptoms instead of causes, approve PRs without reading the callers, and ship Dockerfiles running as root.

Skills fix that. A skill is a folder with a SKILL.md playbook (plus optional scripts) that the agent loads automatically when a task matches. SkillForge packs proven engineering practice into skills that are:

  • 🎯 Opinionated and specific. Concrete steps, commands, checklists and output formats, not vague advice.
  • ✅ Verification-first. Every skill ends with "prove it": run the tests, measure, trace the data flow. No unverified claims.
  • 🛠️ Backed by real tools. Zero-dependency Python scanners for secrets and dangerous code patterns, plus a changelog generator, all unit-tested.
  • 🔌 Portable. Plain SKILL.md folders following the open Agent Skills format. One-command install for Claude Code (as a plugin marketplace), Codex CLI, or any agent that reads skill folders.
  • 🧩 Modular. Five plugins. Install only what you need.

📦 Skill catalog

🧑‍💻 dev-workflow: everyday engineering

Skill What it does
commit-craft Conventional Commits from the real diff; splits mixed work into logical commits
pr-description Reviewer-friendly PRs with testing steps, risk and rollback
code-review Prioritized, verified findings across correctness, security, performance and tests
systematic-debugging Reproduce → isolate → git bisect → root cause → regression test
tdd-workflow Red, green, refactor using your project's existing test stack
safe-refactor Characterization tests first, then small verified behavior-preserving steps

🔐 security: find it before attackers do

Skill What it does
security-audit OWASP Top 10 / CWE audit with a bundled pattern scanner (30+ rules, 8 languages)
secret-scanner Finds leaked keys and tokens in files and git history, then guides rotation
dependency-audit CVE triage by reachability, safe upgrade plans, license and supply-chain checks

📚 docs: documentation people actually read

Skill What it does
readme-architect READMEs that convert visitors into users, with every command verified
changelog-keeper Keep a Changelog + SemVer bump suggestion, generated from Conventional Commits
adr-writer Architecture Decision Records (MADR) with honest trade-offs
api-docs OpenAPI 3.1 specs derived from real route handlers and validators

🚀 devops: ship it safely

Skill What it does
dockerize Multi-stage, non-root, cache-friendly Dockerfiles for Node, Python and Go
github-actions Fast CI with SHA-pinned actions, least privilege, OIDC and injection-safe scripts
incident-postmortem Blameless postmortems with timelines and owned, dated action items

🎨 frontend: fast and accessible

Skill What it does
accessibility-audit WCAG 2.2 AA audit: axe + keyboard, screen reader and contrast checks, fixes in source
web-performance Core Web Vitals (LCP, INP, CLS): measure, fix the biggest bottleneck, re-measure

⚡ Quick start

Option 1: Claude Code plugin marketplace

/plugin marketplace add meyshad/skillforge
/plugin install dev-workflow@skillforge
/plugin install security@skillforge

Available plugins: dev-workflow, security, docs, devops, frontend.

Option 2: one-line installer (any supported agent)

curl -fsSL https://raw.githubusercontent.com/meyshad/skillforge/main/scripts/install.sh | bash
# Pick a target and a subset
./scripts/install.sh --target codex --plugin security     # ~/.codex/skills
./scripts/install.sh --project --skill code-review        # ./.claude/skills in the current repo
./scripts/install.sh --dir ~/my-agent/skills              # any agent that reads SKILL.md folders
./scripts/install.sh --list                               # see everything
./scripts/install.sh --uninstall --plugin frontend

Option 3: manual

Copy any folder from plugins/*/skills/ into your agent's skills directory. Each skill is self-contained.

🧠 How it works

You:   "review my changes before I open the PR"
Agent: matches the `code-review` skill description → loads SKILL.md
       → scopes the diff → runs correctness/security/perf/test passes
       → verifies each finding → reports by severity with concrete fixes

Only the short description of each skill sits in the agent's context. The full playbook loads on demand, so installing all 18 skills costs almost nothing until one is needed. You can also invoke a skill by name: "use the secret-scanner skill on this repo".

🔧 Bundled tools (usable on their own)

# Hard-coded secrets in files and full git history (exit 1 on findings, CI-friendly)
python3 plugins/security/skills/secret-scanner/scripts/scan_secrets.py . --history

# First-pass SAST: eval, shell injection, SQL concatenation, unsafe deserialization, TLS off, ...
python3 plugins/security/skills/security-audit/scripts/scan_patterns.py . --min-severity high

# Changelog section + next SemVer from Conventional Commits
python3 plugins/docs/skills/changelog-keeper/scripts/changelog.py --version auto

Python 3.8+ standard library only. No pip install.

🗂️ Repository layout

.claude-plugin/marketplace.json   # plugin marketplace manifest
plugins/<plugin>/
  .claude-plugin/plugin.json      # plugin manifest
  skills/<skill>/SKILL.md         # the playbook (YAML frontmatter + instructions)
  skills/<skill>/scripts/         # optional helper tools
scripts/install.sh                # cross-agent installer
scripts/validate.py               # structure + frontmatter validator (runs in CI)
template/SKILL.md                 # starting point for new skills
tests/                            # unit tests for bundled tools

🤝 Contributing

New skills and improvements are very welcome. Start from template/SKILL.md, read the skill authoring guide, and run:

python3 scripts/validate.py && python3 -m unittest discover tests

See CONTRIBUTING.md for details. Ideas: open an issue with the Skill request template.

🗺️ Roadmap

  • database plugin: migration review, query optimization, index advisor
  • mobile plugin: React Native / Flutter release checklists
  • ai-engineering plugin: prompt evaluation, RAG debugging
  • Skill evaluation suite (before/after task benchmarks)

⭐ Star history

If SkillForge saves you time, a star helps others find it.

Star History Chart

License

MIT © 2026 Meyshad

About

Agent Skills for serious software engineering: commits, code review, debugging, security audits, Docker & CI.

Topics

Resources

Code of conduct

Contributing

Security policy

Stars

1 star

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages