investigating-gha-failures
Pass
Audited by Gen Agent Trust Hub on Sep 3, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [Indirect Prompt Injection Surface]: The skill fetches and analyzes build logs from GitHub Actions, which represent an external, untrusted data source. If these logs contain malicious instructions, they could potentially influence the agent's behavior during the debugging or code-fixing process.
- Ingestion points: Data enters the context via the output of
tools/fetch_gha_logs.pyas described inSKILL.md. - Boundary markers: There are no explicit delimiters or instructions provided to the agent to ignore embedded commands within the fetched logs.
- Capability inventory: The skill is intended to help the agent diagnose and fix issues in source code, which involves file-write operations.
- Sanitization: No specific filtering or sanitization steps for the log content are described in the instructions.
- [Local Command Execution]: The skill invokes a local Python script
tools/fetch_gha_logs.pyto interact with GitHub APIs and process files. This is a standard pattern for extending agent capabilities using specialized local tools, but it relies on the integrity of the local script.
Audit Metadata