The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides playbooks that perform development and maintenance operations via the shell, including Git worktree management, simulator cleanup (xcrun simctl), and directory removal (rm -rf) for disk space reclamation within the worktree-cleanup playbook.
[COMMAND_EXECUTION]: The workflow integrates with local scripts for project orchestration and maintenance, specifically referencing scripts/orch/orch.ts (executed via Bun) and scripts/worktree-audit.sh.
[COMMAND_EXECUTION]: The Shipping and Autopilot playbooks utilize automated CLI tools for pull request management, such as the Graphite (gt) stack tool.
[SAFE]: The skill implements strong defensive practices against Indirect Prompt Injection. The Babysit and Bugbot Triage references explicitly instruct agents to treat all external comments and review text as "untrusted evidence" and never as executable instructions.
[SAFE]: All irreversible or high-impact actions, including deployments, force-pushing to shared branches, and merging code, are explicitly gated behind mandatory human review checkpoints as detailed in the 'Autonomy and checkpoints' and 'Workflow quality defaults' sections.
[SAFE]: The skill is well-structured for multi-agent coordination across various platforms (Claude Code, Cursor, Codex, etc.) using standardized capability mapping rather than suspicious obfuscation or hidden execution paths.

poteto-mode