issue-report-dev
Pass
Audited by Gen Agent Trust Hub on Oct 15, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses local CLI tools including
gitandghto perform technical investigations and manage issues. These actions are restricted to the local repository and the final issue creation is explicitly gated by user approval. - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted input such as technical logs and stack traces. To mitigate the risk of injection, the skill implements a multi-phase orchestration protocol: 1. Ingestion points: User-provided logs in
report.mdandartifacts/. 2. Boundary markers: Established via the Parent/Worker agent architecture with defined checkpoints for user verification. 3. Capability inventory: Read-only file access,git loghistory analysis, and scoped GitHub write permissions. 4. Sanitization: Handled by a mandatory human-in-the-loop review of the registration plan before any external writes. - [DATA_EXFILTRATION]: Technical analysis is sent to GitHub as part of the intended issue creation flow. This targets a well-known service and requires explicit user consent, aligning with safe practices for developer tools.
Audit Metadata