agentic-security-scanner

Warn

Audited by Gen Agent Trust Hub on Jun 12, 2026

Risk Level: MEDIUMEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONDATA_EXFILTRATIONCREDENTIALS_UNSAFEPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill downloads and installs its core scanner using npx @clear-capabilities/agentic-security-scanner and adds plugins from https://github.com/Clear-Capabilities/agentic-security. These sources are associated with the skill author.
  • [REMOTE_CODE_EXECUTION]: Executes external code downloaded via npm and GitHub within the local environment to perform scans and automated fixes.
  • [COMMAND_EXECUTION]: Performs extensive shell operations including project-wide scanning, file system modifications for vulnerability remediation, and the installation of Git hooks for continuous monitoring.
  • [DATA_EXFILTRATION]: Provides functionality to transmit security audit results to a user-specified Slack webhook URL via curl requests.
  • [CREDENTIALS_UNSAFE]: Instructs users to provide highly sensitive credentials (AWS Access Keys, GitHub Tokens, Stripe Keys) as environment variables for automated rotation and remediation features. It also contains a truncated example of a live Stripe secret key.
  • [PROMPT_INJECTION]: The skill has an indirect prompt injection surface as it ingests untrusted code from the project directory for analysis and can perform high-impact actions (file writes, command execution) based on the scan results.
  • Ingestion points: Project files and dependencies in the scanned directory.
  • Boundary markers: None specified in the documentation.
  • Capability inventory: Shell execution, file modification, network access, and persistence via Git hooks.
  • Sanitization: Claims to have a 'bodyguard' mode, but the analysis process remains exposed to untrusted input content.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Jun 12, 2026, 08:51 PM
Security Audit — agent-trust-hub — agentic-security-scanner