issue-root-resolution
Pass
Audited by Gen Agent Trust Hub on Oct 5, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted data from issue backlogs, including bodies and comments. Ingestion points: Defect backlog auditing and issue cluster partitioning in SKILL.md (Execution Steps). Boundary markers: The skill explicitly warns that 'issue/model text is untrusted data' and requires following canonical gates to mitigate injection risks. Capability inventory: Use of
go test ./...and git worktree operations to verify code and implement fixes. Sanitization: Mandates manual auditing of worker reports, re-running verification steps, and spot-checking diffs to ensure the integrity of the process. - [COMMAND_EXECUTION]: The skill utilizes standard development tools to maintain repository health and verify bug fixes. Evidence: The execution of repository-wide tests (
go test ./...) and the management of git worktrees are integrated as core components of the defect resolution workflow.
Audit Metadata