next-decision
Use only when the user says /next-decision or /next-decision with a decision id (D1, D2, or a number). Do not use while implementing, listing decisions, or discussing a choice in passing.
- License 许可
- MIT
- 所属仓库
- sazlin/loadout
- 最近更新
- 2026-11-14
- Stars
- ★ 0
- Forks
- 1
01 前言 Front Matter
- name
- next-decision
- description
- Use only when the user says /next-decision or /next-decision with a decision id (D1, D2, or a number). Do not use while implementing, listing decisions, or discussing a choice in passing.
02 资源文件 3
- 参考文档evals/evals.json3.0 KB
- 参考文档evals/files/session-decisions-done.md243 B
- 参考文档evals/files/session-decisions.md241 B
03 SKILL.md 正文
Next-decision
Command only. Follows the /decisions skill. Review one listed decision so a human can judge whether it was the right call.
This is not "the next decision to make". While next: is a D-id, listed choices are not "already done" — they still need review. When next: is done, they have already been reviewed.
Steps
- Ensure
.session-decisions.mdexists, then picknext::- If
.session-decisions.mdis missing, run/decisionsfirst, then write the file now. - If this session already has a grok list, reuse it when writing the file instead of requiring a second scan — do not use "has a list" as a reason to skip file creation.
- After that write, pick
next:as in step 2 (next:key missing → D1).
- If
- Pick the target:
/next-decision→ the id innext:(D1 if missing)/next-decision D2or/next-decision 2→ that id
- If the user passed no id and
next:isdone, say every listed decision has been reviewed. Stop. Do not treatdoneas a decision id. - If the list is empty or the id is unknown, say so. Stop.
- Write one review using the recipe below. No other decisions.
- After the review, update
next:if this run consumed the cursor, then leave the file uncommitted:- No id (bare
/next-decision) → setnext:to the id after the one just reviewed, ordoneif none remain. - Explicit
D<n>/n→ leavenext:unchanged unless that id equals the currentnext:value, in which case advance it the same way. - Do not commit
.session-decisions.md.
- No id (bare
Review recipe
## D<n>: <short title>
**Choice:** what was selected
**Alternatives:** what else was on the table, or "none considered"
**Why:** the reason given at the time
**Stakes:** what this affects (files, API, scope) and how reversible it is
**Trade-offs:** costs and risks
**Review point:** one sentence a human can agree or disagree with
Enough to judge the call. Not a session recap. Not a one-line restatement. Not an offer to reverse unless they ask.
Common mistakes
| Temptation | Do this instead |
|---|---|
next: is a D-id, and "Nothing next — those are already done" |
Review the next listed choice |
next: is done and the user passed no id |
Say every listed decision has been reviewed. Stop. |
Treat /next-decision as future work |
Review a past choice from the list |
| Dump every decision in full | One id this run |
| One-line restatement | Full recipe |
| Trigger on "we decided X" mid-task | Command only |
04 同仓库更多 Skills
demo
Demo skill
writing-skills
Use when creating new skills, editing existing skills, or verifying skills work before deployment
writing-plans
Use when you have a spec or requirements for a multi-step task, before touching code
verification-before-completion
Use when about to claim work is complete, fixed, or passing, before committing or creating PRs - requires running verification commands and confirming output before making any success claims; evidence before assertions always
using-superpowers
Use when starting any conversation - establishes how to find and use skills, requiring skill invocation before ANY response including clarifying questions
using-git-worktrees
Use when starting feature work that needs isolation from current workspace or before executing implementation plans - ensures an isolated workspace exists via native tools or git worktree fallback