split-stack

Pass

Audited by Gen Agent Trust Hub on Aug 13, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill installs the gh-stack extension using the command 'gh extension install github/gh-stack'. This targets the official repository of a well-known service and is a trusted source for development tooling.- [COMMAND_EXECUTION]: Executes various Git and GitHub CLI commands to manage local and remote branches. It also runs a provided Python script 'strip_hunks.py' to automate code modifications. The script includes safe handling of shell commands via subprocess without shell expansion for environment discovery.- [COMMAND_EXECUTION]: The 'strip_hunks.py' script implements rigorous path validation in its 'resolve' function. It explicitly refuses absolute paths, directory traversal using '..', and symlinks, ensuring all file operations are confined to the repository root.- [PROMPT_INJECTION]: The skill processes external pull request data which presents a surface for indirect prompt injection. This is addressed through a mandatory evidence chain: 1) Ingestion points: Pull request diffs, titles, and bodies processed in Step 1. 2) Boundary markers: Explicit instructions for the classification subagent to return only structured JSON and perform no actions. 3) Capability inventory: File modification through a local script and branch management via Git. 4) Sanitization: Employs a mandatory verification step in Step 4 that requires a zero-byte diff against the original commit SHA to guarantee code integrity.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 13, 2026, 05:50 PM
Security Audit — agent-trust-hub — split-stack