skill-improver
Pass
Audited by Gen Agent Trust Hub on Mar 17, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes the
ghq rootcommand to determine the base directory for git repositories on the local system, which is used to resolve paths for skill source code.\n- [PROMPT_INJECTION]: The skill processes unvalidated user feedback to directly modify the instructions (SKILL.md) of other skills, creating an indirect prompt injection surface.\n - Ingestion points: User feedback provided in the conversation context (e.g., "Improve this skill").\n
- Boundary markers: Absent; the skill is instructed to interpret feedback and apply changes without specific delimiters or warnings to ignore embedded instructions.\n
- Capability inventory: Execution of the
ghqcommand and file-system write access to skill definitions in~/.claude/skills/and local repository paths.\n - Sanitization: Absent; the skill relies on the agent's natural language interpretation to incorporate user feedback into the skill's core logic.
Audit Metadata