investigate-bug
Pass
Audited by Gen Agent Trust Hub on Sep 29, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONNO_CODE
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied data through the
$ARGUMENTSvariable without using boundary markers or delimiters. This allows potentially malicious instructions within a bug report to influence the agent's behavior. - Ingestion points: The
$ARGUMENTSplaceholder inSKILL.mddirectly interpolates external input into the prompt. - Boundary markers: None present; the input is concatenated directly after the "Investigate bug:" prefix.
- Capability inventory: The skill mentions invoking a "consultant CLI" and a "consultant agent" to perform root cause analysis and state analysis.
- Sanitization: No evidence of input sanitization or filtering of the
$ARGUMENTScontent. - [NO_CODE]: The skill consists primarily of natural language instructions and metadata without providing the source code for the referenced 'consultant CLI' or associated scripts, limiting the ability to audit the underlying execution environment.
Audit Metadata