The Agent Skills Directory

[SAFE]: The skill is purely instructional and does not include any scripts or executable code. It defines strict operational boundaries (Read-only, Artifact-only, State-changing) that restrict the agent's ability to modify files or run commands.
[SAFE]: No external network requests, data exfiltration patterns, or third-party dependencies were detected. The skill relies on local project documentation and user-supplied context.
[SAFE]: The instructions include robust safeguards against hallucination and misinformation by explicitly forbidding the invention of reasons, goals, outcomes, or security/performance claims.
[SAFE]: The skill uses clear, non-obfuscated language and focuses on high-quality, durable documentation practices such as ADRs (Architecture Decision Records) and findings reports.
[SAFE]: The skill provides explicit guidance on preserving the meaning and constraints of source text, which helps mitigate potential indirect prompt injection by ensuring the agent remains a faithful editor rather than an obedient executor of embedded instructions.

vibe-writing