The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill fetches gitignore templates from the well-known gitignore.io service (hosted by Toptal). This is standard and expected behavior for repository maintenance.
[COMMAND_EXECUTION]: The skill utilizes common development tools such as git, fd, and rg to inspect the repository state and cp to create a backup of the existing .gitignore file before modification.
[DATA_EXFILTRATION]: To retrieve relevant templates, the skill sends a comma-separated list of detected project technologies (e.g., 'node', 'typescript') to the gitignore.io API. This information is project metadata rather than sensitive personal or credential data.
[PROMPT_INJECTION]: The skill processes local file names and git status output as input, which represents a potential surface for indirect prompt injection.
Ingestion points: Repository file names (via fd) and the output of git status (processed in SKILL.md).
Boundary markers: None explicitly used when presenting data to the agent, although the workflow requires user confirmation for empirical patterns.
Capability inventory: Writing to the .gitignore file and executing shell commands.
Sanitization: The skill filters filenames against a hardcoded detection table in references/LANGUAGE-DETECTION.md before using them as API keys, and requires user confirmation for empirical patterns, which significantly reduces the injection surface.

setup-gitignore