# Verdict

> Say what to do from any input, such as a doc, draft, issue, PR or idea, with pushback, an outcome per item and a chip for each open question, or judge one Plate or Plite review scope by its exact id with a Stop, Pursue or Defer written into its review page's front matter. A scope verdict the owner types also grades the scope against other editors and picks its next step to build, cut or redesign, and `/loop verdict <scope id>` keeps building those steps. `explore <scope or idea>` settles a direction with the owner first. It restates the goal, then runs rounds of pushback and research that each stop for the owner's answer, and gives no verdict until the owner says `lock`. A bare `audit` gives the source files no scope owns to scopes and cards that verdict for the next three scopes no review has ranked, and `audit restart` makes every closed scope due again.

- Skill: `udecode/verdict` (Agent Skill)
- Install (CLI): `npx get-skill add udecode/verdict`
- Raw SKILL.md: https://api.skillmd.com/api/skills/udecode/verdict/raw
- Safety review: pending (external: skill-scanner PASS, skillspector PASS)
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: udecode (https://skillmd.com/u/udecode)
- Updated: 2026-11-15
- Page: https://skillmd.com/skills/udecode/verdict

---


# Verdict

Handle $ARGUMENTS.

Run the Verdict playbook, `.agents/playbooks/verdict.md`, in full: its Paths section picks intake, the deep path or Explore from the argument, or one step of its Audit pass for a bare `audit` or `audit restart`, Explore restates the owner's goal and stops before any research, then runs one round for each answer from the owner until `lock`, intake gives each item of the input one outcome and offers a chip for each open one, and the deep path reconciles earlier reviews, compares the design space with the `best-api` lens, ranks a named scope's next steps by comparing the scope with other editors and writes one Stop, Pursue or Defer into the review page's front matter. Load `pstack:poteto-mode` first, follow the steps of the pstack playbook the Verdict playbook `extends` with its deltas applied, and run `check-playbooks.mjs` as `AGENTS.md`'s How work runs says, because this entry skips poteto-mode's routing. Its Sessions section titles the session with the input or scope first and leaves later work to chips.

