Skip to main content

decisions

Decision journal and retrospectives. Lists one-way door decisions the user made, and replays ones that are due for review (what we expected vs what happened) to build calibrated judgment.

ソース情報

リポジトリ
pipeshub-ai/sharp
ソースの最終更新活動
2026年10月15日 19:46
検出された SKILL.md の言語
英語
スター
4
フォーク
0

インストール方法

デフォルトでは、最初にソースを確認する Prompt が選択されています。直接コマンドに切り替えるか、ローカルコピーをダウンロードすることもできます。

ソースファイルを確認

インストールを決める前に、SKILL.md と SkillsMP に表示されている付属ファイルをお読みください。

SKILL.md を表示中

SKILL.md
ソースの指示 · 読み取り専用プレビュー
name
decisions
description
Decision journal and retrospectives. Lists one-way door decisions the user made, and replays ones that are due for review (what we expected vs what happened) to build calibrated judgment.
argument-hint
[list | review]
disable-model-invocation
true
allowed-tools
Bash(node ~/.sharp/bin/sharp.mjs *)
# /sharp:decisions Decisions due for a retrospective: ```! node ~/.sharp/bin/sharp.mjs decision due 2>/dev/null || echo "[]" ``` If `$ARGUMENTS` is `list`, show `node ~/.sharp/bin/sharp.mjs decision list` as a compact table (date, decision, choice, reason, outcome) and stop. Otherwise, for each due decision, one at a time: 1. Remind them: the decision, **the context at the time**, the options, what they chose, **their reason in their own words**, and the pre-mortem. 2. Ask: "What actually happened?" Then: "Knowing only what you knew *then*, in that context, was it a good decision? Has the context changed since, so that the same call would be wrong today?" Keep **decision quality** separate from **outcome luck**: a good decision can have a bad outcome, and the reverse. 3. Ask for the lesson in one line, then record it: ```bash node ~/.sharp/bin/sharp.mjs decision review <id> --outcome "…" --verdict "good-call|bad-call|good-call-bad-luck|lucky" --lesson "…" ``` 4. If the lesson generalizes, save it as a lesson (`lesson add --when … --principle …`). Judgment improves only with feedback. This closes the loop that one-way doors usually leave open. ## How to write - Plain words, for a smart newcomer. Define each technical term in a few words the first time you use it. - Short paragraphs. Finish every part you start; never end on a fragment. - Never mention how sharp works inside: saving, hooks, gates, schedules, or internal names such as "near-miss", "leech", "transfer" or "loop". - Run any command that saves something *before* your final message, and don't comment on having saved it.
GitHubで見る