The Agent Skills Directory

[SAFE]: The skill implements a robust safety contract that requires the agent to confirm metered operations, obtain separate exact approval for deletions, and report if required tools are missing from the environment.
[SAFE]: Approval boundaries are clearly defined, explicitly prohibiting the expansion of credit budgets, unauthorized contact with customers, or the deletion of reusable assets.
[SAFE]: No evidence of sensitive file access, network exfiltration, prompt injection, or obfuscation was found in the skill instructions or configuration.

buildbetter-synthetic-research