The Agent Skills Directory

[SAFE]: The skill implements a safety protocol for 'Tier-2' actions such as database migrations, file deletions, and security configuration changes. It mandates a detailed plan, a risk assessment, and a mandatory 'STOP' point to wait for explicit user confirmation before any execution occurs.
[SAFE]: The use of an 'Adversarial Critique' step encourages the agent to internally verify and challenge its own plan for potential failures before presenting it to the user.
[SAFE]: The skill includes verification steps and abort conditions to ensure that the agent stops immediately if a step fails or if the environment state changes unexpectedly during execution.

gravitas-highstakes