woostack-execute

Pass

Audited by Gen Agent Trust Hub on Sep 7, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSMETADATA_POISONING
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses shell commands for worktree management, Git/Graphite repository operations, and executing focused verification/smoke tests within isolated environments. These are restricted to the repository root and scoped worktree paths.
  • [EXTERNAL_DOWNLOADS]: Fetches authoritative project and issue specifications from trusted providers (Linear, Plane, GitHub) and repository metadata from Git/Graphite. These operations serve the primary purpose of state verification and delivery.
  • [METADATA_POISONING]: Evaluates the risk of deceptive instructions in processed project metadata (Category 7). It mitigates this by using an 'assume-malicious' posture where instructions in issue titles, descriptions, or comments are treated as data and never allow the skill to bypass its own safety boundaries (e.g., merge prohibitions).
  • [SAFE]: The skill implements strong security boundaries: mutually exclusive input modes, deterministic task selection, immutable ordinals, and mandatory no-follow directory traversal for local files. It explicitly defines that it never performs merges, reviews, or marks product acceptance.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 7, 2026, 03:42 PM
Security Audit — agent-trust-hub — woostack-execute