plugin-update

Pass

Audited by Gen Agent Trust Hub on Oct 8, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses the Bash tool to execute system discovery and update commands.\n
  • Executes claude plugin commands to list installed plugins, fetch marketplace details, and perform installations or updates.\n
  • Runs local utility scripts scripts/discover-plugins.sh and scripts/fetch-latest-versions.sh to handle complex logic for version checking and configuration parsing.\n- [EXTERNAL_DOWNLOADS]: The skill retrieves metadata and installation packages from the author's GitHub repository.\n
  • Fetches plugin.json manifests from raw.githubusercontent.com/kochetkov-ma/claude-brewcode to verify the latest available versions.\n
  • Uses the gh tool to query the GitHub API for latest release tags.\n
  • Configures a third-party plugin marketplace pointing to https://github.com/kochetkov-ma/claude-brewcode.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes external data that could theoretically influence agent behavior, though it is used here for legitimate version management.\n
  • Ingestion points: Reads and parses plugin.json files from the local plugin cache and remote GitHub repository.\n
  • Boundary markers: None explicitly defined in the helper scripts for parsing the version strings.\n
  • Capability inventory: The skill has access to Bash, Write, and WebFetch tools, providing a wide surface if external metadata were manipulated.\n
  • Sanitization: The logic uses jq to strictly parse specific JSON fields, which mitigates the risk of the agent interpreting data as instructions.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 8, 2026, 06:27 PM