push-skill-to-github

Pass

Audited by Gen Agent Trust Hub on Sep 7, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill utilizes the cmux tool to execute shell commands in a background terminal pane.
  • It performs repository-wide staging using git add -A and executes git commit and git push within the ~/.agents directory.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes skill files that may contain instructions designed to influence the agent during the staging or review process.
  • Ingestion points: The skill interacts with any files located in ~/.agents/skills (via SKILL.md).
  • Boundary markers: The instructions explicitly direct the agent to wait for user requests and verify changes before pushing.
  • Capability inventory: The skill has shell execution access (cmux) and network access via the Git protocol (git push).
  • Sanitization: The skill does not perform automated content sanitization, instead relying on the user's manual review and a claimed external "sanitized public mirror" process.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 7, 2026, 03:30 AM
Security Audit — agent-trust-hub — push-skill-to-github