gsd-pr-branch

Warn

Audited by Gen Agent Trust Hub on Apr 3, 2026

Risk Level: MEDIUMCOMMAND_EXECUTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill delegates its entire logic to an external workflow file located at @/mnt/local-analysis/workspace-hub/.codex/get-shit-done/workflows/pr-branch.md. This pattern of dynamic instruction loading hides the actual shell commands and agent behaviors from the skill's main definition file.
  • [COMMAND_EXECUTION]: The mapping logic in the <codex_skill_adapter> section explicitly enables complex agent orchestration tools like spawn_agent and request_user_input, which are used to execute the external workflow's tasks without explicit visibility in the primary file.
  • [PROMPT_INJECTION]: The skill has a surface for indirect prompt injection as it processes git commit data and branch history (untrusted input) to perform its filtering objective.
  • Ingestion points: Git commit messages and file diffs processed during the filtering logic (SKILL.md).
  • Boundary markers: Absent; there are no instructions to ignore embedded commands within the git data.
  • Capability inventory: Uses spawn_agent for multi-agent orchestration and implies git CLI usage (SKILL.md).
  • Sanitization: Absent; the skill does not define any sanitization or validation for the data ingested from the git environment.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Apr 3, 2026, 04:33 AM