Where decisions live and who makes them - open questions, closed decisions, rejected proposals - so the team never re-argues a settled point or decides something that is the founder's. Use when a decision is needed, taken, deferred or challenged.
Installs into .claude/skills of the current project.
Are you the author of Decision Log?
Add the live security badge to your README. It updates with every re-scan.
[](https://www.skillsdirectory.com/skills/fadi88-decision-log)
---
name: decision-log
description: "Where decisions live and who makes them - open questions, closed decisions, rejected proposals - so the team never re-argues a settled point or decides something that is the founder's. Use when a decision is needed, taken, deferred or challenged."
---
<!-- Generated by tools/sync_adapters.py from skills/decision-log/SKILL.md. Edit the source, not this file. -->
# Decision log
Three files in `docs/requirements/`:
- **`p1-open-questions.md`:** what is waiting on the founder. Each entry: the question, the
options, your recommendation, what it blocks, the date raised.
- **`p1-closed-decisions.md`:** what was decided. Each entry: date, decision, reason, who
decided, what would make it worth reopening.
- **`p1-rejected.md`:** proposals the founder turned down, with why, recorded the day they are
turned down, whoever proposed them (an agent, a review, a draft the founder pasted in). Only
what was actually put to the founder: an option the template itself rules out (an uncapped free
tier, say) is not a founder's rejection. Reopening one needs a new
fact, not a fresh opinion.
## What is the founder's
Targets, thresholds, prices, scope, the stack, legal positions, risk acceptance, permissions,
known reds, and anything outward. Agents bring the numbers and a recommendation; the founder
decides; the decision is written down the same day.
## Before proposing anything
Grep `p1-closed-decisions.md` and `p1-rejected.md`. If it was settled, say so and stop, unless
you have a new fact; then say what the new fact is.
## One event is not an era
Do not change a rule, a model or a target because of one odd result. Explain the odd result.
Change the rule when the pattern repeats.