The Agent Skills Directory

[DATA_EXPOSURE]: The skill is instructed to read potentially sensitive files such as .env, credentials, and local configurations to build a project model. However, it includes a strict safety rule: "Never embed secret values found in .env, credentials, keys, tokens, or local config," which mitigates the risk of accidental credential exposure in public documentation.
[PROMPT_INJECTION]: As a documentation tool, the skill processes untrusted content from the software project (manifests, existing READMEs, and source code). This presents an indirect prompt injection surface where a malicious project could attempt to influence the agent's behavior. The risk is significantly reduced by the workflow requirement to "Show per-file diffs and wait for confirmation" before writing any files, ensuring a human reviews the generated output.
[COMMAND_EXECUTION]: The skill explicitly restricts its own behavior to static analysis for its first version, stating: "Do not run the target project unless the user asks; static analysis is enough for v1." This prevents the accidental execution of malicious code within the project being documented.

doc-gen