diagnosing-bugs
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to create and run various diagnostic commands, including shell scripts, CLI invocations, and network requests via
curlto establish a feedback loop for bug reproduction. These actions are standard for developer-oriented tools and are confined to the user's local development environment. - [INDIRECT_PROMPT_INJECTION]: The skill involves analyzing external, potentially untrusted data such as application logs, network traces, and core dumps. This creates a potential surface for indirect instructions to be introduced into the agent's context.
- Ingestion points: Artifacts captured during Phase 1 (traces, logs) and Phase 4 (instrumentation logs).
- Boundary markers: The skill includes a mandatory
Redactsection to strip secrets from outputs, helping to isolate sensitive information from the reasoning process. - Capability inventory: The agent utilizes shell execution, file system access, and network operations to perform its diagnostic tasks.
- Sanitization: The instructions specifically mandate the removal of credentials, environment variables, and authentication headers from any data presented or processed.
Audit Metadata