debugging-assistant
Pass
Audited by Gen Agent Trust Hub on Feb 17, 2026
Risk Level: SAFE
Full Analysis
- Indirect Prompt Injection (LOW): This skill exposes a vulnerability surface for indirect prompt injection by processing external data with high-privilege capabilities. • Ingestion points: Step 1 in
SKILL.mdinstructs the agent to examine untrusted error messages, stack traces, and logs. • Boundary markers: Absent; the skill lacks instructions for using delimiters or delimiters to separate untrusted data from system prompts. • Capability inventory: Step 2 inSKILL.mdprovides the agent with access to system tools including Bash, Grep, Glob, and Read. • Sanitization: Absent; the workflow does not include instructions to sanitize or validate the content of error logs before analysis.
Audit Metadata