gen-catalog

Pass

Audited by Gen Agent Trust Hub on Mar 23, 2026

Risk Level: SAFEPROMPT_INJECTIONDATA_EXFILTRATIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: Vulnerability to indirect prompt injection through processing of external skill definitions.
  • Ingestion points: Reads SKILL.md and command files from ~/.claude/skills/ and ~/.claude/commands/.
  • Boundary markers: Absent; instructions content is scanned without isolation to determine categories.
  • Capability inventory: File system writes to README and JSON files; git commit and git push operations.
  • Sanitization: Absent; no validation or escaping of parsed skill descriptions before inclusion in output.
  • [DATA_EXFILTRATION]: Potential exposure of internal configuration data via automated version control.
  • The skill accesses ~/.claude/projects/ to record telemetry data.
  • Automated git push in Phase 5 transmits the generated catalog (including names and paths of all installed skills) to a remote repository.
  • [COMMAND_EXECUTION]: Automates shell commands for git management.
  • Executes staging, committing, and pushing of files to an external git remote.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 23, 2026, 10:58 AM
Security Audit — agent-trust-hub — gen-catalog