agent-creator
Use when creating expert agents. Generates agent.md with frontmatter, hooks, required sections, and skill references.
Use when running ecosystem-specific vulnerability scans across project dependencies (npm/composer/pip/cargo/go/etc), with optional auto-fix.
Claude Code
$ npx -y skills add fusengine/claude-plugins --skill dependency-audit --agent claude-code
Codex
$ npx skills add fusengine/claude-plugins --skill dependency-audit -a codex -y
Cursor
$ npx skills add fusengine/claude-plugins --skill dependency-audit -a cursor -y
Gemini CLI
$ npx skills add fusengine/claude-plugins --skill dependency-audit -a gemini-cli -y
GitHub Copilot
$ npx skills add fusengine/claude-plugins --skill dependency-audit -a github-copilot -y
OpenCode
$ npx skills add fusengine/claude-plugins --skill dependency-audit -a opencode -y
Windsurf
$ npx skills add fusengine/claude-plugins --skill dependency-audit -a windsurf -y
name: dependency-audit description: "Use when auditing a project's dependency tree for known vulnerabilities and supply-chain risks (bun/npm/pnpm/yarn, Composer, Cargo, Go modules, SwiftPM, pip, Bundler) and its GitHub Actions workflows. Read-only: reports findings, lists fix commands as options, never applies them." argument-hint: "[ecosystem]" user-invocable: true
<objective> This skill audits a project's dependencies with each ecosystem's native audit tool (detected from lockfiles), adds supply-chain checks (OWASP Top 10:2025 A03 Software Supply Chain Failures) and GitHub Actions workflow checks, and writes one report. It is read-only: it never installs a tool, never runs a fix command, never edits a lockfile. Fix commands are listed in the report as options for the owner.
Out of scope: deep CVE history of one named package (use cve-research). </objective>
1. **Report, never fix.** Every `fix` / `--fix` / `--apply-fixes` command rewrites lockfiles, manifests or `node_modules`. The audit lists it as an option; it never runs it. 2. **Tool presence first.** Before any command: `command -v <tool>`. Missing tool → record `not run: <tool> absent` in the report. Never install it, never skip it silently. 3. **Untrusted projects.** Some tools execute project code (Cargo build scripts, package manager lifecycle scripts, osv-scanner `fix`). Use the read-only form documented in `references/audit-commands-systems.md` and never run install commands. 4. **Versions age.** Check a tool's version with its own `--version` rather than trusting a number written in this skill.
| Lockfile | Native tool | Reference | |----------|-------------|-----------| | `bun.lock` | `bun audit` | [JS and PHP commands](references/audit-commands.md) | | `package-lock.json` / `npm-shrinkwrap.json` | `npm audit` | same | | `pnpm-lock.yaml` | `pnpm audit` | same | | `yarn.lock` (Yarn 2+) | `yarn npm audit` | same | | `yarn.lock` (Yarn 1 classic) | `yarn audit` | same | | `composer.lock` | `composer audit` | same | | `Cargo.lock` | `cargo audit` | [Rust, Go, Swift, Python, Ruby](references/audit-commands-systems.md) | | `go.mod` / `go.sum` | `govulncheck` | same | | `Package.resolved` (SwiftPM) | none native — OSV / GitHub Advisory lookup | same | | `requirements.txt`, `pyproject.toml`, `pylock.*.toml` | `pip-audit` | same | | `Gemfile.lock` | `bundle-audit` | same | | any supported lockfile | `osv-scanner` (optional cross-check) | same |
1. **Detect** lockfiles (and whether each is tracked: `git ls-files <lockfile>`). 2. **Check tools** with `command -v`; record absent ones. 3. **Run** the read-only audit command per ecosystem, JSON output where available. 4. **Supply chain** checks: [Supply Chain Checks](references/supply-chain.md). 5. **CI** checks for `.github/workflows/*.yml`: [GitHub Actions Checks](references/ci-github-actions.md). 6. **Classify** each advisory (CRITICAL / HIGH / MEDIUM / LOW, CWE, confidence 1-10). 7. **Verify** each candidate finding in a second read-only pass that tries to refute it: is the affected version the one actually resolved in the lockfile, is the vulnerable code reachable (govulncheck does this for Go; elsewhere grep imports/calls), is it a dev-only dependency. Confidence below 8 goes to "needs manual review", not to findings. 8. **Write** the report with [Report Template](references/templates/audit-report.md) to `docs/security/YYYY-MM-DD-<scope>.md` inside the audited repository.
Use the advisory's own severity when it states a CVSS score, and record the CVSS version and source (GitHub, NVD, CNA). Tool labels map as: `critical` → CRITICAL, `high` → HIGH, `moderate`/`medium` → MEDIUM, `low`/`info` → LOW. These are inputs: the reported severity follows the plugin's single scale in `rules/security-rules.md`, after reachability. For CVSS details and reachability rules see the cve-research skill.
| Ecosystem | Option for the owner | What it changes | |-----------|---------------------|-----------------| | bun | `bun audit fix` (`--dry-run` to preview, `--latest` to cross majors) | `bun.lock`, `node_modules`, exact pins in `package.json` | | npm | `npm audit fix` (`--dry-run`, `--package-lock-only`; `--force` allows SemVer-major) | `package-lock.json`, `node_modules` | | pnpm | `pnpm audit --fix` (overrides) or `--fix=update` | `pnpm-workspace.yaml` or lockfile | | Cargo | `cargo audit fix` (`--dry-run`); needs `cargo install cargo-audit --features=fix` | `Cargo.toml` | | pip | `pip-audit --fix` (`--dry-run` to preview) | installed environment | | Composer, Go, SwiftPM, Bundler | manual version bump in the manifest, then the package manager's update | manifest + lockfile |
A plugin ecosystem that turns Claude Code into a supervised, multi-agent development environment.
Use when creating expert agents. Generates agent.md with frontmatter, hooks, required sections, and skill references.
Use when starting ANY development task -- feature, bug fix, refactor, hotfix (triggers: implement, create, build, fix, add…
Use when creating a feature/component or adding functionality. Fires BEFORE APEX Analyze to refine requirements via structured…
Use before a root-cause, done/verified claim, irreversible action, or 2nd-time fix reaches the owner (APEX or plain…
Use when validating code quality after modifications -- SOLID compliance, DRY duplication, linter errors, architecture…
Use when an expert agent self-reviews and self-corrects code after the Execute phase, before sniper validation (BMAD-METHOD…
You MUST use this before any creative work - creating features, building components, adding functionality, modifying behavior, or…
Ask which skill or flow fits your situation. A router over the skills in this repo.