skill-lookup
Fail
Audited by Gen Agent Trust Hub on Feb 17, 2026
Risk Level: HIGHEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS] (HIGH): The skill retrieves metadata and file contents (including scripts) from prompts.chat, which is not a trusted source. There is no mechanism to verify the integrity or safety of these downloads.
- [REMOTE_CODE_EXECUTION] (HIGH): By saving downloaded scripts and helper files to the
.claude/skills/directory, the skill creates a direct path for the installation and potential execution of arbitrary code provided by an external source. - [PROMPT_INJECTION] (LOW): The skill has a surface for indirect prompt injection (Category 8) as it processes external content without validation.
- Ingestion points: Data returned by
search_skillsand file contents fromget_skill. - Boundary markers: Absent; the instructions do not define delimiters or warnings to ignore embedded commands within the downloaded content.
- Capability inventory: Local file writing and expansion of agent capabilities.
- Sanitization: Absent; content is saved directly to disk without escaping or verification.
Recommendations
- AI detected serious security threats
Audit Metadata