The Agent Skills Directory

[SAFE]: The skill uses a secure 'path-allocate' pattern for task creation, where task definitions are written to scratch files and then validated and committed. This prevents multi-line content from being passed directly through the shell boundary, mitigating command injection risks.
[SAFE]: Input validation is comprehensive. Subcommands validate task numbers, deliverables, domains, profiles, and skills (bundle:skill format). It also enforces that steps in non-verification tasks are valid file paths.
[SAFE]: The qgate-mechanical-checks component performs deterministic checks to ensure the implementation plan is structurally sound, verifying that every deliverable has a task, dependencies are acyclic, and file paths exist as expected for their intended operations.
[SAFE]: The pre-commit-verify-freshness utility ensures that the codebase state matches the recorded build history by comparing working-tree hashes, providing a robust safety check against pushing unverified code changes.
[SAFE]: The code is modular and transparent, using standard Python libraries. There are no external dependencies on untrusted third-party packages, no network exfiltration patterns, and no hidden or obfuscated code blocks.

manage-tasks