The Agent Skills Directory

[SAFE]: The skill incorporates a mandatory data redaction framework that prevents the exposure of credentials, private file paths, and personally identifiable information (PII) in generated artifacts.\n- [SAFE]: Instructions strictly prohibit the use of unbacked superlatives or marketing language, ensuring that the generated documentation remains factual and reduces the likelihood of hallucinated safety claims.\n- [SAFE]: Tool access is limited to safe local operations such as reading project files and executing standard test commands, with no functionality for unauthorized network access or external script downloads.\n- [SAFE]: The skill's design mitigates indirect prompt injection risks by using rigid templates and focusing on objective, verifiable reporting, which limits the influence of malicious content that might be present in the repository being documented.

write-for-ai