code-debug
Pass
Audited by Gen Agent Trust Hub on Oct 17, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied data in the form of bug descriptions or regressions. This input is used to guide the agent's actions, including command execution, which creates an attack surface for instructions embedded within the user data.
- Ingestion points: The user-provided
<bug or regression>argument defined in theargument-hintofSKILL.md. - Boundary markers: There are no explicit delimiters or instructions provided to the agent to ignore potential instructions embedded within the user-provided bug description.
- Capability inventory: The skill instructs the agent to execute commands to reproduce symptoms (
agent-runnable command), use debuggers, and run regression tests. - Sanitization: The instructions do not specify any validation or sanitization steps for the input before it is used to generate or execute commands.
Audit Metadata