manage-knowhow-capture
Pass
Audited by Gen Agent Trust Hub on Jul 3, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes the
Bashtool to execute system commands, specificallycurl, to fetch content for thereferenceknowledge type. - [EXTERNAL_DOWNLOADS]: The
referencecapture mode (Step 3) downloads content from arbitrary user-provided URLs via the--sourceflag. - [PROMPT_INJECTION]: The skill is susceptible to Indirect Prompt Injection because it ingests and processes untrusted data from external websites.
- Ingestion points: External content is fetched from URLs provided in the
$ARGUMENTSthrough the--sourceflag (SKILL.md). - Boundary markers: The instructions do not specify any delimiters or warnings (e.g., "ignore embedded instructions") to isolate the fetched content from the agent's control flow.
- Capability inventory: The skill possesses
Bashexecution,Writeaccess to the filesystem, and the ability to interact with the user viarequest_user_input(SKILL.md). - Sanitization: There is no mention of sanitization, validation, or escaping of the content fetched via
curlbefore the agent extracts key points from it.
Audit Metadata