The Agent Skills Directory

[SAFE]: The skill mandates a strict three-phase workflow which requires explicit user confirmation ('proceed') before moving from the proposal phase to implementation, serving as a critical safety gate.- [SAFE]: Explicit instructions are provided to manage secrets securely in .env files and to never hardcode or log credentials.- [SAFE]: The 'read before write' rule and the requirement for minimal changes reduce the likelihood of the agent introducing unintended logic or overreaching its assigned scope.- [SAFE]: The system uses a localized, file-based tracking mechanism for tasks and plans, which ensures that all agent actions are documented in the repository for human audit and multi-agent coordination.- [SAFE]: Instructions specifically warn the agent to mark risky commands with explicit warning comments, promoting safer terminal usage during development.

pma