prd-worktree

Pass

Audited by Gen Agent Trust Hub on Aug 13, 2026

Risk Level: SAFECOMMAND_EXECUTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes a bundled shell script (create.sh) to perform Git operations. The script is well-structured, using set -euo pipefail for safety and double-quoting variables to prevent word splitting or command injection. It uses standard utilities like git, find, sed, and tr for its operations.- [INDIRECT_PROMPT_INJECTION]: The skill reads external data from PRD files in the prds/ directory to infer branch names.
  • Ingestion points: Files matching prds/*.md.
  • Boundary markers: Absent.
  • Capability inventory: File system reads, Git worktree creation, and local file copying (cp).
  • Sanitization: The skill performs significant sanitization on the extracted PRD titles, converting them to URL-safe slugs using tr and sed to remove any potentially malicious characters before using them in shell commands.- [PRIVILEGE_ESCALATION]: The skill computes paths for new worktrees outside the current repository directory (../). While this is standard practice for Git worktrees to avoid nested repositories, it is a capability that involves creating directories in the parent folder.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 13, 2026, 05:40 PM
Security Audit — agent-trust-hub — prd-worktree