bug-fix
Pass
Audited by Gen Agent Trust Hub on May 6, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: Indirect prompt injection surface detected in the bug-fixing workflow.
- Ingestion points: The skill ingests untrusted data from user-provided bug reports via the
$ARGUMENTSvariable inSKILL.mdand from files within the codebase during investigation. - Boundary markers: Absent. Prompts in
AGENTS.md(e.g., Tracer and Fixer prompts) interpolate user-derived summaries directly into task instructions without protective delimiters or instructions to ignore embedded commands. - Capability inventory: The skill possesses significant capabilities including the ability to launch sub-agents with the
Agenttool and to modify the filesystem in Stage 4 ofSKILL.md. - Sanitization: No sanitization or escaping is applied to bug reports or codebase content before they are used to construct instructions for the sub-agents.
Audit Metadata