The Agent Skills Directory

[DYNAMIC_CONTEXT_INJECTION]: The SKILL.md file uses the !command syntax to execute scripts/gitlab_context.py at skill load time. This script performs network operations and local environment scanning before user interaction.
[DATA_EXPOSURE]: The script scripts/validate_glfm.py attempts to read the user's ~/.bashrc file to extract GITLAB_TOKEN. Accessing shell configuration files is a sensitive operation that exposes the contents of the user's environment profile.
[COMMAND_EXECUTION]: Multiple scripts (get_gitlab_context.py, gitlab_context.py) use subprocess.run or gitpython to execute shell commands (glab, git). While used for context gathering, this provides a mechanism for arbitrary command execution if arguments are improperly handled.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from .gitlab-ci.yml files and GitLab API responses (pipeline status, commit messages) via gitlab_context.py without explicit sanitization or boundary markers. This creates a surface where malicious content in a repository could influence agent behavior.
[EXTERNAL_DOWNLOADS]: The scripts/sync_gitlab_docs.py script downloads a documentation archive from gitlab.com and replaces local files. Although the source is a well-known service, the automated download and extraction of remote content into the skill's reference directory is a significant capability.
[CREDENTIALS_UNSAFE]: The references/gitlab-ci-local-guide.md file contains placeholder credentials like glpat-YOUR_TOKEN_HERE. While these are placeholders, the skill's design encourages users to store sensitive tokens in local configuration files (~/.gitlab-ci-local/variables.yml) which the skill's scripts then access.

gitlab-skill