fix-traceback
Fail
Audited by Gen Agent Trust Hub on Feb 16, 2026
Risk Level: HIGHPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [Indirect Prompt Injection] (HIGH): The skill creates a direct path for untrusted data to influence file system modifications and shell execution.
- Ingestion points: External tracebacks and error logs (SKILL.md).
- Boundary markers: Absent; no delimiters are defined to separate the data being analyzed from the agent's instructions.
- Capability inventory: The skill allows the agent to 'implement the smallest fix' (file write/edit) and 'provide a verification command' (command execution).
- Sanitization: Absent; no validation or escaping of the input log content is specified.
Recommendations
- AI detected serious security threats
Audit Metadata