file-test-bug
Pass
Audited by Gen Agent Trust Hub on Sep 16, 2026
Risk Level: SAFEDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTION
Full Analysis
- Local File Access: The skill reads junit.xml, agent-metadata.md, and test source code to identify and diagnose failures. This access is local and specific to the tests/reports directory and source files mentioned in test reports.\n- Data Transmission: The skill is designed to send local log contents to a GitHub repository. While the target repository belongs to a trusted organization, it is worth noting that the skill requests the complete, unmodified logs for inclusion in the issue body, which may include details of the agent's internal operations during test runs.\n- Indirect Data Processing: The skill ingests data from external test reports and metadata to generate a diagnosis. Every Cat 8 finding requires documenting the evidence chain: (1) Ingestion points include junit.xml and agent-metadata.md files. (2) There are no explicit boundary markers or instructions to ignore embedded content in the logs. (3) The skill possesses the capability to create GitHub issues using the github-mcp-server-create_issue tool. (4) The skill includes a step for the agent to analyze and diagnose the root cause before reporting, which acts as a logical check on the data.
Audit Metadata