issue-sweep
Pass
Audited by Gen Agent Trust Hub on Aug 15, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill implements advanced safety protocols for autonomous operations. It uses a combination of advisory GitHub labels (agent:processing) with TTL and deterministic branch naming conventions (claude/issue-) to coordinate multiple agent instances and prevent duplicate Pull Requests. A categorical 'Never auto-merge' policy ensures that all automated changes undergo human review prior to integration.
- [COMMAND_EXECUTION]: The skill executes git, gh, and bun commands for repository maintenance and issue tracking. These are limited to the intended workflow, using specific environment variables like AGENTLOOP_WORKTREE_BASE to manage temporary files safely and preventing unauthorized system-wide modifications.
- [PROMPT_INJECTION]: The skill manages a significant surface for indirect prompt injection by processing GitHub issues. It mitigates this via strict boundary markers (identity headers and sweep-trace HTML comments) and logic to distinguish machine-authored reports from human directives. High-risk actions, such as security-sensitive fixes or architectural changes, are automatically escalated for human confirmation.
Audit Metadata