Back to Neuro OS Playbook
Self-Check Skill
Source: playbook-source-package/skills/self-check/SKILL.md
Self-Check
A mandatory verification skill for worker agents. Ensures every agent checks their own work before marking a task as done.
Description
The Self-Check is the first layer of verification in the two-layer quality system. Every worker agent must run this checklist before marking any task as "done" — no exceptions, not even for "small changes."
Procedure
Generic checks (all roles)
- [ ] The deliverable exists at the expected location
- [ ] The deliverable is accessible (can be opened/loaded/viewed)
- [ ] The deliverable matches what was requested in the task description
- [ ] No placeholder content or TODO markers remain
- [ ] I have not recreated something that already existed (check PROJECT-INVENTORY.md)
Role-specific checks
Customize per agent role. The check must be concrete and verifiable — not "code works" but "code runs without errors, all acceptance criteria met, no secrets committed."
Anti-patterns tracking
Maintain a living table of past mistakes:
| Anti-Pattern | What Went Wrong | Check |
|---|---|---|
| [Added as issues arise] | [Description] | [What to verify] |
A mistake that happens once is a bug. Twice is a process failure. Three times means the instructions are not clear enough — rewrite them.
Applicable Roles
- All worker agents
Tags
- verification
- self-check
- mandatory