decision-journal
Skill4 starsChanged 4 months agoMITView on GitHub
Appends structured entries to a decisions.md journal — context, options considered, the choice, a two-failure-mode premortem, and a review-by date — and runs reviews when decisions come due. Use when the user says "log this decision", "I decided to...", "add to my decision journal", "review my decisions", or is weighing a significant choice worth recording. Don't use for daily habit check-ins — use habit-tracker — or for weekly work retrospectives — use weekly-review.
5 other projects use a version of this.
Adopted by
Most recently taken up by
- ohanyan/adultingtheir own version, first seen Nov 16, 2026
- crissmoldovan/agent-skillstheir own version, first seen Oct 15, 2026
- ba0918/claude-skillstheir own version, first seen Jul 9, 2026
- Xipher-Labs/walter-ostheir own version, first seen May 21, 2026
- itseffi/agentic-ostheir own version, first seen Jan 23, 2026
Other versions5 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 5
- itseffi/agentic-os★ 114same name
- Xipher-Labs/walter-os★ 8same name
- crissmoldovan/agent-skills★ 2same name
- ohanyan/adulting★ 1same name
- ba0918/claude-skills★ 0same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository74 other files it gives its agents.
- agent-eval-harnessSkill
- context-budget-managerSkill
- incident-rollbackSkill
- memory-hygieneSkill
- model-cost-optimizerSkill
- prompt-debuggerSkill
- skill-author-coachSkill
- third-party-skill-vetterSkill
- api-endpoint-testerSkill
- bug-triage-protocolSkill
- changelog-release-notesSkill
- docs-readme-writerSkill
- env-secrets-hygieneSkill
- github-pr-reviewSkill
- git-ship-flowSkill
- lovable-brief-writerSkill
- supabase-migration-writerSkill
- vercel-deploy-checkSkill
- cold-outreach-sequencerSkill
- competitor-pricing-scanSkill
- contract-redline-scanSkill
- customer-support-inboxSkill
- invoice-runnerSkill
- kpi-dashboard-weeklySkill
- lead-research-dossierSkill
- payment-reconciliationSkill
- pipeline-trackerSkill
- proposal-builderSkill
- stripe-payment-linkSkill
- testimonial-harvesterSkill
- weekly-revenue-reportSkill
- brand-voice-guardianSkill
- case-study-builderSkill
- content-repurposerSkill
- deep-research-briefSkill
- email-nurture-sequenceSkill
- landing-page-copywriterSkill
- linkedin-content-packSkill
- newsletter-assemblerSkill
- seo-blog-writerSkill
- social-listening-reportSkill
- x-thread-builderSkill
- youtube-script-writerSkill
- budget-guardrailsSkill
- ceo-operating-doctrineSkill
- delegation-protocolSkill
- editor-personaSkill
- heartbeat-standupSkill
- hiring-briefSkill
- ops-agent-personaSkill
- researcher-personaSkill
- sales-agent-personaSkill
- support-agent-personaSkill
- weekly-retroSkill
- writer-personaSkill
- calendar-conciergeSkill
- email-triageSkill
- file-organizerSkill
- habit-trackerSkill
- inbox-zero-sprintSkill
- The first 60 shown.
The file
Files
Decision Journal
Builds the highest-leverage feedback loop a founder can own: a plain decisions.md where every significant choice is recorded before the outcome is known — context, options, choice, a premortem of the top two failure modes, and a date to come back and score it. The standard: entries are written at decision time (no retro-fitting), the journal is append-only, and due reviews actually happen.
When to Use
- The user just made — or is about to make — a decision with real stakes (money, direction, people, commitments).
- "Log this decision", "add to the journal", "I'm going with option B because...".
- Review time: "any decisions due for review?", "score my past calls".
- Not for: trivial reversible choices (don't journal lunch), habit tracking (
habit-tracker), or general weekly reflection (weekly-review).
Quick Reference
| Action | Command / Call |
| Locate the journal | ls decisions.md in the user's notes folder (create with a # Decision Journal header if missing) |
| Next entry ID | grep -c "^## D-" decisions.md then +1 |
| Append an entry | write the filled template to the END of decisions.md — never edit old entries |
| Find due reviews | python3 due-scan (Procedure step 5) |
| Today / review-by math | date +%F · python3 -c "from datetime import date,timedelta; print(date.today()+timedelta(days=90))" |
Procedure
- Precheck — locate
decisions.md; create with header if missing (confirm location with the user first time). Determine the mode: logging a new decision or reviewing due ones. - Gather (logging mode) — pull from the user, prompting only for what's missing: the context (what situation forced a choice, in 2–3 sentences with the key numbers), the real options considered (minimum two — "do nothing" counts and often should be listed), and the choice with the primary reason. Write in the user's words; this is their judgment on record, not the agent's.
- Premortem — exactly two failure modes — ask or propose: "It's {review date} and this decision went badly. The two most likely reasons?" Each failure mode gets one line plus an early-warning sign to watch for. Two, not five — forcing the top two is what makes the premortem sharp.
- Set review-by and append — default review horizons: 30 days (tactical), 90 days (most business decisions), 180+ (strategic); compute the date with python3. Fill the template, assign the next
D-{n}ID, and append to the end of the file. Show the user the entry before writing; append only on their OK. Never modify prior entries — corrections become new dated notes under the entry. Review mode — scan for due entries:
python3 - <<'EOF' import re, datetime as dt text = open('decisions.md').read() today = dt.date.today() for m in re.finditer(r'^## (D-\d+): (.+?)$(.*?)(?=^## D-|\Z)', text, re.M | re.S): body = m.group(3) due = re.search(r'Review by: (\d{4}-\d{2}-\d{2})', body) reviewed = 'Outcome:' in body if due and not reviewed and dt.date.fromisoformat(due.group(1)) <= today: print(f"DUE {m.group(1)}: {m.group(2)} (review by {due.group(1)})") EOFFor each due entry: reread it with the user, then append a dated Outcome block — what happened, verdict (good call / bad call / good process-bad luck / bad process-good luck), whether either premortem failure mode fired, and the one lesson. The process/luck distinction is the point of the whole journal.
- Deliver — confirm what was appended (
tail -20 decisions.md) and surface any still-due reviews so they don't rot.
Output Template
## D-{n}: {Decision title}
Date: {YYYY-MM-DD} · Review by: {YYYY-MM-DD} · Status: open
**Context:** {2–3 sentences, key numbers included}
**Options considered:**
1. {Option A} — {one-line case for it}
2. {Option B / do nothing} — {one-line case}
**Choice:** {what was decided} — because {primary reason}.
**Premortem (top 2 failure modes):**
1. {Failure mode} — early sign: {observable warning}
2. {Failure mode} — early sign: {observable warning}
<!-- appended at review time -->
**Outcome ({date}):** {what happened}. Verdict: {good call / bad call / good process, bad luck / bad process, good luck}.
Premortem hit? {1 / 2 / neither}. Lesson: {one sentence}.Pitfalls
- Entry written after the outcome is known — hindsight contaminates the record and the journal becomes a highlight reel. Recovery: log at decision time; if backfilling an old decision, mark it
(backfilled — hindsight risk)so future-you discounts it honestly. - Premortem sprawls to five vague risks — "it might not work" teaches nothing. Recovery: exactly two failure modes, each concrete enough to have an observable early-warning sign; if a risk has no observable sign, sharpen it until it does.
- Reviews never happen — the journal fills, the loop never closes, no learning. Recovery: run the due-scan at every journal touch (and during
weekly-review); a due entry blocks the "all clear" report until reviewed or explicitly deferred with a new date. - Outcome graded on results alone — a lucky bad process gets reinforced. Recovery: the verdict vocabulary forces the process/luck split; if the verdict line says only "it worked", it isn't done.
- Old entries edited to look smarter — never; the journal is append-only. Recovery: corrections and updates are new dated lines under the entry.
Verification
- Entry appended at file end; all prior entries byte-identical (no edits above)
- Options ≥2, premortem = exactly 2 failure modes, each with an early-warning sign
- Review-by date computed, in the future, and ISO-formatted
- Due-scan run this session; due entries reviewed or explicitly re-dated
- Entry shown to the user before writing; wording is theirs
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

