diagnosing-bugs
Pass
Audited by Gen Agent Trust Hub on Oct 13, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill focuses on standard software development practices such as unit testing, regression testing, and log analysis. It explicitly instructs the agent to protect user privacy by redacting secrets and sensitive data from any captured artifacts or command outputs.
- [COMMAND_EXECUTION]: The skill requires the agent to execute local shell commands, tests, and scripts (e.g.,
curl,git bisect,bash) to reproduce and diagnose bugs within the user's development environment. These operations are necessary for the skill's primary function and are performed under user oversight. - [INDIRECT_PROMPT_INJECTION]: The skill involves processing external data sources such as application logs, source code, and network traces. While these inputs are theoretically untrusted, the skill incorporates strong defensive instructions, requiring the agent to redact sensitive information and maintain a high level of scrutiny before processing external data.
- Ingestion points: Project source code, architecture records (ADRs), application logs, HAR files, and user-provided inputs from the human-in-the-loop script.
- Boundary markers: Specific instructions are provided to use
<REDACTED>placeholders for secrets and to quote only relevant portions of captured logs. - Capability inventory: Execution of local shell commands, file system reads for codebase exploration, and potential browser automation for UI testing.
- Sanitization: Mandatory redaction of credentials and authorization headers prior to displaying or using captured data.
Audit Metadata