error-debugging-error-trace
Pass
Audited by Gen Agent Trust Hub on Oct 7, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDATA_EXFILTRATIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The ErrorTrackingAnalyzer script defined in resources/implementation-playbook.md performs recursive file reads across the project directory (Path(project_path).rglob('*')). This creates an ingestion point where malicious instructions embedded in the codebase (e.g., in comments or strings) could be interpreted by the agent during the analysis phase.
- Ingestion points: ErrorTrackingAnalyzer._analyze_error_handling in resources/implementation-playbook.md reads content from all project source files.
- Boundary markers: There are no instructions or delimiters provided to the agent to distinguish between its instructions and the content of the files it reads.
- Capability inventory: The skill provides Python scripts capable of reading filesystem content and processing it via regex and AST parsing.
- Sanitization: No sanitization or filtering is applied to the content read from the codebase before the agent processes it.
- [DATA_EXFILTRATION]: The skill provides implementation patterns for transmitting application logs and error data to external endpoints, including Sentry, Slack (via webhooks), and custom HTTP APIs. While this is the intended purpose of the skill and the templates include sanitization logic to redact secrets (e.g., in ErrorTracker.sanitizeEvent), it establishes a mechanism for data to leave the local environment.
- [COMMAND_EXECUTION]: The instructions direct the agent to assess error handling using provided analysis scripts. This involves the execution of Python code to traverse and read the local filesystem to identify patterns in the user's codebase.
Audit Metadata