agentsec is a zero-config CLI that audits every skill your AI agent runs — against the OWASP Agentic Skills Top 10. Agent-platform agnostic: supports Claude Code, OpenClaw / ClawHub, Codex / skills.sh, Hermes (Nous Research), and generic project-local skill directories.

One command. Full security report. No sign-up.

Quick Start

npx agentsec

No flags needed. agentsec walks every default skills directory on your machine — grouped by platform — plus any ./skills folder in the current project (up to two levels deep).

Example Output

  ✔ Found 7 skills
  ℹ Scanned 5 locations across Claude Code, OpenClaw, Codex / skills.sh, Hermes, Other
    Claude Code
      ~/.claude/skills (3 skills)
    OpenClaw
      ~/.openclaw/workspace/skills (1 skill)
    Codex / skills.sh
      ~/.agents/skills (1 skill)
    Hermes
      ~/.hermes/skills (1 skill)
    Other
      ./skills (1 skill)

  Claude Code (3 skills)
  ✔ fetch-data     v1.0.0  D (42)
  ✔ deploy-helper  v2.3.0  C (68)
  ✔ code-review    v1.1.0  A (95)

  OpenClaw (1 skill)
  ✔ summarize-docs v0.9.0  A (91)

  Codex / skills.sh (1 skill)
  ✔ db-migrate     v1.4.2  B (78)

  Hermes (1 skill)
  ✔ research-assistant v0.3.1  B (82)

  Other (1 skill)
  ✔ lint-fix       v2.0.0  A (93)

  7 skills scanned  •  avg score 79  •  5 certified
  Findings: 2 critical, 1 high, 2 medium

  ⚠ WARN  3 high/critical finding(s) detected
  Run with --verbose for detailed findings and recommendations.

Auto-discovery

npx agentsec with no arguments scans every default skills directory for every agent platform it supports:

| Platform | Paths scanned | | ---------------------- | ------------------------------------------------------------------------------------------------------------------------- | | Claude Code | ~/.claude/skills, ./.claude/skills, ~/.claude/plugins/*/skills/*, ~/.claude/commands, ./.claude/commands | | OpenClaw / ClawHub | ~/.openclaw/workspace/skills, ~/.openclaw/workspace-*/skills (profiles via OPENCLAW_PROFILE), ~/.openclaw/skills | | Codex / skills.sh | ~/.agents/skills, ./.agents/skills, ../.agents/skills, /etc/codex/skills | | Hermes | ~/.hermes/skills, ~/.hermes/skills/* (bundled categories), ./.hermes/skills | | Other (generic) | Any skills/ directory inside the current project, up to two levels deep |

Pass --path <dir> to audit a specific directory instead, or --platform <claude\|openclaw\|codex\|hermes> to narrow to one platform.

Installation

# Run ad-hoc (recommended)
npx agentsec

# Or install globally
npm install -g agentsec
bun install -g agentsec

Commands

# Audit every default skills directory on this machine
agentsec

# Show detailed findings, score breakdowns, and recommendations
agentsec --verbose

# Scan a specific path
agentsec scan --path ./my-project

# Apply a policy preset
agentsec audit --policy strict
agentsec audit --policy enterprise

# Narrow to one platform (openclaw | claude | codex | hermes)
agentsec --platform claude

# Output formats: text | json | sarif | html
agentsec --format json --output audit.json
agentsec --format sarif --output audit.sarif
agentsec --format html --output report.html

# Generate a report from a saved JSON audit
agentsec report audit.json --format html --output report.html

# Inspect policy presets
agentsec policy list

Run agentsec help for the full option reference.

OWASP Agentic Skills Top 10

Every scan checks all 10 risk categories from the OWASP AST10:

| ID | Risk | What We Detect | | --------- | ----------------------- | ------------------------------------------------------- | | AST01 | Malicious Skills | Dangerous code patterns, known-malicious signatures | | AST02 | Supply Chain Compromise | Dependency provenance, transparency log gaps | | AST03 | Over-Privileged Skills | Excessive permission grants, least-privilege violations | | AST04 | Insecure Metadata | Schema validation failures, metadata integrity issues | | AST05 | Unsafe Deserialization | Parser safety gaps, injection vectors | | AST06 | Weak Isolation | Missing sandboxing, container misconfigurations | | AST07 | Update Drift | Unpinned versions, stale dependencies, hash mismatches | | AST08 | Poor Scanning | Coverage gaps, incomplete scanning pipelines | | AST09 | No Governance | Missing audit logs, absent policy enforcement | | AST10 | Cross-Platform Reuse | Platform-specific validation gaps, portability issues |

Supported Agent Platforms

agentsec is agent-platform agnostic — every platform that emits an agentskills.io SKILL.md is auditable.

• Claude Code — scans installed skills, plugins, and MCP servers
• OpenClaw / ClawHub — full SKILL.md manifest analysis, including metadata.openclaw.* extensions
• Codex / skills.sh — skill and plugin scanning under ~/.agents/skills and /etc/codex/skills
• Hermes — Nous Research's Hermes Agent, including metadata.hermes.* extensions (tags, category, requires_toolsets, config)
• Generic — any project-local ./skills directory

Output Formats

| Format | Flag | Best For | | ------ | ---------------- | ------------------------------------- | | Text | --format text | Terminal, human-readable (default) | | JSON | --format json | CI pipelines, programmatic processing | | SARIF | --format sarif | VS Code, GitHub Code Scanning | | HTML | --format html | Stakeholder reports, dashboards |

CI/CD Integration

Fail your build when a scan finds high-severity issues:

# Exit 1 on any high or critical finding
agentsec audit --policy strict --format sarif --output audit.sarif

Upload the resulting SARIF to GitHub Code Scanning to track findings over time.

Requirements

• Bun >= 1.0.0 — the CLI is built and distributed as a Bun bundle.

Links

• 🌐 Website: agentsec.sh
• 📦 GitHub: semiotic-ai/agentsec
• 🐛 Issues: github.com/semiotic-ai/agentsec/issues
• 🧩 Skills ecosystem: skills.sh

License

MIT © AgentSec Contributors