graphite-skill
Pass
Audited by Gen Agent Trust Hub on Feb 17, 2026
Risk Level: SAFE
Full Analysis
- COMMAND_EXECUTION (SAFE): The skill includes shell scripts for installation (
install.sh) and verification (test/verify-installation.sh). These scripts are well-documented, transparent, and perform standard local configuration tasks within the project's.claudedirectory without unauthorized or malicious behavior. - PROMPT_INJECTION (SAFE): Through the
hooks/session-start.shscript, the plugin injects instructions into the agent's context to optimize task delegation and output formatting. While this directs agent behavior, it is the primary intended function of the skill and does not attempt to bypass safety constraints or security filters. - DATA_EXPOSURE (SAFE): The skill interacts solely with local Git and Graphite metadata. No evidence of hardcoded credentials, sensitive file access (such as SSH keys or environmental variables), or unauthorized network exfiltration was found.
- REMOTE_CODE_EXECUTION (SAFE): All scripts provided are local. The skill does not perform remote downloads, nor does it use piped-to-shell patterns (e.g.,
curl | bash) during installation or runtime.
Audit Metadata