file-test-bug
Pass
Audited by Gen Agent Trust Hub on Feb 17, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- Indirect Prompt Injection (LOW): The skill reads external, potentially untrusted data from local test reports and log files, then includes this content verbatim in a GitHub issue and uses it for automated diagnosis.
- Ingestion points: The skill reads
tests/reports/junit.xml(Step 2) andagent-metadata.md(Step 4). - Boundary markers: Absent. The instructions explicitly require the "complete, unmodified contents" and "verbatim" inclusion of logs.
- Capability inventory: The skill uses
github-mcp-server-create_issueto send data to an external repository. - Sanitization: Absent. There is no escaping or filtering of the content read from the local files before it is processed by the agent or posted to GitHub.
- Data Exposure (SAFE): While the skill reads local files and sends them to the internet, the destination is a trusted GitHub repository (
microsoft/github-copilot-for-azure) and the data is limited to build artifacts (test results and execution metadata) relevant to the skill's stated purpose.
Audit Metadata