# cisco-ai-defense/skill-scanner

Security Scanner for Agent Skills

Repository: https://github.com/cisco-ai-defense/skill-scanner
Canonical: https://ross.abutalabs.com/products/skill-scanner
Homepage: https://cisco-ai-defense.github.io/docs/skill-scanner
Language: Python
License: NOASSERTION
License Family: other
Topics: agent, agent-skills, security
Last push: 2026-08-04T04:50:35+00:00

## Health v2 (maintenance only)
Score: 80/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 96, release rhythm 96, longevity 15
- inputs: {"age_days": 217, "days_push": 29, "days_rel": 30, "gap_med": 7, "n_releases_24m": 18}
- flags: no_license
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 2460, forks 312 (observed 2026-08-28T04:06:53.401346+00:00)

## What it is
A security scanner for AI agent skills that detects prompt injection, data exfiltration, and malicious code patterns using pattern-based detection (YAML + YARA), LLM-as-a-judge analysis, and behavioral dataflow analysis. It supports OpenAI Codex Skills and Cursor Agent Skills formats, integrates with CI/CD via SARIF output, and offers a plugin architecture for custom analyzers.

## Use cases
- scan agent skills for prompt injection before installing them
- detect data exfiltration in AI agent skill packages
- find malicious code patterns in Claude Code or Codex skills
- add security scanning of agent skills to a GitHub Actions pipeline
- audit skill repositories for hardcoded secrets and obfuscation
- check agent skills for vulnerable dependencies via OSV.dev
- run pre-commit scans on skill markdown files

## When to choose
- you install or distribute AI agent skills and want to vet them for threats
- you need CI/CD integration with SARIF output for GitHub Code Scanning
- you want layered detection combining static analysis, behavioral dataflow, and LLM review
- you need to scan multiple skill formats including Codex, Cursor, and Claude Code commands

## When to avoid
- you need certified or guaranteed security - the scanner is explicitly best-effort only
- you want to scan general-purpose codebases rather than agent skill packages
- you need runtime protection of a live agent rather than pre-deployment scanning

## Facets
- artifact type: cli-tool
- maturity: active
- function: security, vulnerability-scanning, static-site-generator, cli, llm-inference, developer-tools
- domain: security, artificial-intelligence, developer-tools
- platform: python, cli, cross-platform
- tags: agent-skills, prompt-injection, yara, sarif, static-analysis, llm-as-a-judge, dataflow-analysis, pre-commit, ci-cd, supply-chain-security, ai-agents, command-line

## Member repositories
- cisco-ai-defense/skill-scanner (main) score 80

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:06:53.401346+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T02:29:19.978520+00:00, confidence not recorded.
  - readme: https://github.com/cisco-ai-defense/skill-scanner (fetched 2026-08-28T04:06:53.401346+00:00, sha c6d076422ab7)
  - homepage: https://cisco-ai-defense.github.io/docs/skill-scanner (fetched 2026-08-29T10:11:19.186978+00:00, sha 9099672cf03a)
- Data as of 2026-08-30T08:39:29.467469+00:00.