check-skill-conflicts
Fail
Audited by Gen Agent Trust Hub on Feb 17, 2026
Risk Level: HIGHEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- EXTERNAL_DOWNLOADS (HIGH): The skill directs the agent to execute a shell script from the 'plinde-plugins' marketplace. This source is untrusted and the code being executed is not verified, posing a significant risk of executing unvetted third-party code.
- COMMAND_EXECUTION (MEDIUM): Multiple shell commands are suggested for direct execution, including script invocation and destructive file operations like 'git rm' and 'mv' on local configuration directories.
- PROMPT_INJECTION (MEDIUM): The skill acts as an Indirect Prompt Injection vector by processing 'SKILL.md' files from external sources (plugins). Maliciously crafted skill metadata could potentially influence the conflict checker's output or the agent's resolution decisions.
- DATA_EXPOSURE (LOW): The skill accesses the user's private agent configuration directory ('~/.claude/skills/'), which may contain proprietary or sensitive logic, although this access is consistent with its stated purpose.
Recommendations
- AI detected serious security threats
Audit Metadata