The Agent Skills Directory

[SAFE]: The skill instructions promote safety and accuracy by requiring agents to provide proof of success. No malicious behaviors such as data exfiltration, prompt injection, or unauthorized command execution were identified. The mentioned commands (e.g., pytest, npm test, jq, terraform) are provided as illustrative examples of standard verification practices. No external dependencies or remote script executions are present. The instruction to independently verify subagent work and check version control diffs serves as a defensive measure against following potentially incorrect or malicious automated suggestions.

verification-before-completion