The Agent Skills Directory

[SAFE]: The skill implements a structured workflow with mandatory 'Phase Gates', requiring explicit user confirmation before moving from planning to code modification. This ensures the user remains in control of the agent's actions at every critical step of the development process.
[SAFE]: Command execution is strictly localized to the project's existing scripts (such as npm test, pytest, or make lint) for verification purposes. The skill documentation explicitly instructs the agent to use the project's own scripts rather than raw or arbitrary system tools.
[SAFE]: The skill maintains persistence of project state locally within the .absolute-work/board.md file. It does not attempt to create hidden persistence mechanisms or modify system configuration files outside the defined project scope.
[SAFE]: No evidence of data exfiltration, obfuscation, or malicious prompt injection was found. The instructions emphasize 'Codebase-First Intelligence', directing the agent to search the local environment for configuration and facts before querying the user, which reduces unnecessary interaction and aligns with least-privilege information gathering.
[SAFE]: The verification framework includes a 'Generator-Evaluator' separation, which uses a separate subagent or context to review implementation quality, providing an internal layer of defense against logic errors and accidental security regressions.

absolute-work