skills-manager

Pass

Audited by Gen Agent Trust Hub on May 3, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill instructions involve executing multiple local shell scripts (e.g., skill-activation-prompt.sh, add-keywords-to-skill.sh, extract-keywords.sh) located in the .claude/hooks/ directory. It also includes instructions to modify file permissions using chmod +x to ensure these local automation hooks are executable.
  • [PROMPT_INJECTION]: The skill describes an 'Auto-Learning Feature' where user input from the prompt is processed to extract keywords. These keywords are then automatically injected into the local skill-rules.json configuration via a shell script. This creates a surface for indirect prompt injection where adversarial user input could potentially influence the agent's skill-routing logic.
  • Ingestion points: User prompts (processed in SKILL.md).
  • Boundary markers: None specified.
  • Capability inventory: Execution of shell scripts and writing to local JSON configuration files (add-keywords-to-skill.sh).
  • Sanitization: No explicit sanitization or validation of the extracted keywords is documented.
Audit Metadata
Risk Level
SAFE
Analyzed
May 3, 2026, 02:14 AM