The Agent Skills Directory

[SAFE]: The skill is primarily documentation-based. The manifest.json and SKILL.md (frontmatter) explicitly define tools: [], meaning the agent cannot execute any code through this skill.
[COMMAND_EXECUTION]: While scripts/_skill_cli.py contains a subprocess.run call in a helper function, this script is not exposed as a tool in the manifest. It appears to be a boilerplate utility common in this vendor's skill sets and is not active in this specific skill.
[PROMPT_INJECTION]: The instructions in SKILL.md and context/ANTI_PATTERNS.md (e.g., "Always run...", "Never ask...") are task-specific behavioral constraints designed to enforce a specific CI workflow. They do not attempt to bypass core AI safety filters or jailbreak the model.
[DATA_EXPOSURE]: The skill mentions an official support email (support@heyeddi.com) and standard GitHub PR workflows. No sensitive credentials, tokens, or private file paths are exposed or accessed.

heyeddi-ci-guide