Decision doc
Skill 0xelitesystem/claude-skills-templates/skills/decision-doc
Five production-ready Claude Skill templates: one-pager writer, code reviewer, brand voice keeper, CSV cleaner, decision-doc generator. Each demonstrates a different SKILL.md pattern. Drop into Claude Code or Claude.ai. MIT licensed.
npx -y skills add 0xelitesystem/claude-skills-templates --skill decision-docAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
One thing to look at
- 1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
What its author says it does
Copied from the file, not written here
Produce a structured decision document with explicit tradeoffs, weighted criteria, scored options, a recommendation, and a flip condition. Use whenever the user is choosing between options, comparing approaches, evaluating vendors or frameworks, or deciding between architecture patterns. Triggers on "help me decide", "compare X vs Y", "which should I pick", "should I use A or B", framework or vendor decisions, architecture choices, or any decision where a structured tradeoff would be more useful than a casual opinion. Always use this skill rather than free-form pros and cons lists when the user is making a real decision.
SKILL.md
3.3 KB, as published. Nobody here has run it
Decision Doc
Produces a structured decision document. Forces a recommendation with confidence and a flip condition, instead of hedging.
When to use
- Two or more options under consideration
- The user wants a recommendation, not just a comparison
- The decision has stakes (money, time, lock-in, reversibility)
When NOT to use
- The user is exploring a topic without a specific decision in front of them.
- One option has a hard disqualifier. Just say so, no need for the full doc.
- The user explicitly asks for a casual opinion.
Process
Step 1: Restate the decision
One sentence in your own words. Flag any ambiguity that would change the answer. If ambiguity is severe, ask one clarifying question before proceeding.
Step 2: Define criteria
List 4 to 7 evaluation criteria. Assign each a weight from 1 to 5. Total weights must sum to a number between 15 and 25. This constraint prevents flattening the criteria with all-equal weights.
If the user provided criteria, use those. If not, derive them from the decision context.
Step 3: Score each option
Score each option from 1 to 5 against each criterion. Compute a weighted total per option. Show the work.
Step 4: State the tradeoff for each option
One sentence per option describing what you give up by choosing it. No option escapes this step. Even the best option has a cost.
Step 5: Recommend
State the recommendation. Cite the weighted score, but do not let the score override clear qualitative concerns. If a lower-scoring option is the right answer for a reason the score cannot capture, say so.
Step 6: Flip condition
State one specific change in inputs or context that would make a different option win. This is the most important field. If you cannot articulate a flip condition, the analysis is not done.
Step 7: Confidence
1 (coin flip) to 5 (clear winner).
Output format
## Decision
<restated decision>
## Ambiguities
<list, or "none">
## Criteria
| Criterion | Weight | Why it matters |
|---|---|---|
## Scoring
| Option | <criterion 1> | <criterion 2> | ... | Weighted Total |
|---|---|---|---|---|
## Tradeoffs
- <Option A>: <what you give up>
- <Option B>: <what you give up>
## Recommendation
<option name>
<2 to 4 sentences explaining why>
## Flip Condition
If <specific input change>, the recommendation becomes <other option>.
## Confidence
<1 to 5> / 5
Stability check
For high-stakes decisions, run the analysis twice with the options listed in opposite order. If the recommendation flips between runs purely from ordering, the analysis is unstable. Widen the criteria or get more information before committing.