The Agent Skills Directory

[SAFE]: The skill consists entirely of Markdown-based documentation providing technical guidance, architectural patterns, and historical lessons. No malicious executable code, active scripts, or hidden payloads are included.
[DATA_EXPOSURE]: The reference files contain extensive lessons on protecting sensitive data. For example, they provide specific patterns for preventing cross-tenant leaks in multi-tenant APIs, masking secrets in diagnostic logs, and handling PII in mobile applications. No hardcoded credentials or exfiltration patterns were found.
[PROMPT_INJECTION]: The instructions are designed to guide the agent through technical review loops (e.g., adjudicating findings, managing test quality). No attempts to override safety filters, bypass system instructions, or use adversarial role-play were detected.
[EXTERNAL_DOWNLOADS]: The documentation references external open-source projects for adaptation and learning (e.g., 'delegate-skills' and 'planning-with-files'). These are documented neutrally as case studies and target established platforms like GitHub.
[COMMAND_EXECUTION]: Various shell commands and Python snippets are included as templates for the agent to use during development tasks (e.g., database migrations, CI configuration). These are informational in nature and not designed for automatic or malicious execution.

review-lessons