codify
Skill0 starsChanged 4 months agoMITView on GitHub
Capture a just-made decision into one durable record so the reasoning survives the session. Writes a single markdown file to the project's decisions/ folder: the call, the reasoning, the weakest assumption, what would change the verdict, and when to reuse the pattern. Checks for an existing record on the same decision and updates it instead of creating a near-duplicate. Activates on "codify this decision", "record this decision", "save the decision record", "make this decision durable", or right after a kill-go verdict, decision-brief, or /decide session.
7 other projects use a version of this.
Adopted by
Most recently taken up by
- helmut-hoffer-von-ankershoffen/heinzeltheir own version, first seen Nov 8, 2026
- vic-lsh/agent-skillstheir own version, first seen Nov 8, 2026
- DoKu67/claude-code-skilstheir own version, first seen Oct 16, 2026
- yuanhaorannnnnn/agent-skillstheir own version, first seen Sep 18, 2026
- treekey/claude-codifytheir own version, first seen Jul 6, 2026
- anombyte93/atlas-ai-skillstheir own version, first seen Apr 26, 2026
Other versions7 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 7
- anombyte93/codify★ 4same name
- DoKu67/claude-code-skils★ 4same name
- helmut-hoffer-von-ankershoffen/heinzel★ 2same name
- vic-lsh/agent-skills★ 2same name
- treekey/claude-codify★ 1same name
- anombyte93/atlas-ai-skills★ 0same name
- yuanhaorannnnnn/agent-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 repository15 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- ai-graduationSkill
- arbitrage-auditSkill
- assumption-mapperSkill
- b2b-idea-vettingSkill
- decision-briefSkill
- design-thinking-liteSkill
- grill-meSkill
- grill-with-docsSkill
- innovation-diagnosticSkill
- kill-goSkill
- preflightSkill
- problem-framingSkill
- pull-frameworkSkill
The file
Files
Codify
Use this immediately after a decision is made — a kill-go verdict, a finished decision-brief, the end of a /decide session, or any call the user just committed to. The job is to make the decision durable: one file, one decision, written so a reader six months from now understands what was decided, why, and when to reuse the pattern.
Clarity compounds only if it's captured. A decision that lives in chat history is a decision the team will re-litigate.
The Write Procedure
- Anchor the location. Records go in
decisions/at the root of the user's project repo — not the current working directory if that's a subdirectory, and never inside this plugin. Createdecisions/on first use. - Slug the decision. Derive a short kebab-case filename from the decision itself:
decisions/kill-freemium-tier.md,decisions/pivot-to-eval-service.md. - Check before creating — always. List
decisions/and read the closest-matching existing record before writing anything. (No directory-listing tool available? Ask the user to name any existing record on this decision — do not skip the check silently.) If an existing record covers the same decision (same call being made or revisited, even under a different name), update that file: revise the fields that changed and add a dated addendum line under Context. Create a new file only when no record covers this decision. Near-duplicate records are the failure mode this step exists to prevent — if the folder accumulates two records for one decision, that drift is the signal this check needs tightening. - Fill the template honestly. Record the actual confidence and the genuinely weakest assumption, not a sanitized version. If the decision was a kill, say kill and say why. The record is for truth, not for looking decisive.
The Record Template
# Decision: [the call, stated as a decision, not a topic]
Date: [YYYY-MM-DD]
## Context
[2-3 sentences: what prompted this decision, what was at stake]
## The Call
[What was decided. One or two sentences, no hedging.]
## Reasoning
[Why this call over the alternatives. The strongest case that won.]
## Weakest Assumption
[The assumption most likely to be wrong, stated plainly]
## What Would Change This
[Specific evidence or events that would reopen the decision]
## Reuse This Pattern When
[The situations where this decision's logic applies again]
## Confidence
[High / Medium / Low — matched to evidence quality, not enthusiasm]Discoverability Check
After writing the record, check whether the project's AGENTS.md or CLAUDE.md points agents at decisions/. If neither file mentions it, propose the one-line addition to the user — state the exact line in your response and ask before writing it. For example: "Durable decision records live in decisions/ — read the relevant record before revisiting a past decision." Never silently edit the user's context files; they own them. If neither file exists, say so in one line and move on.
Anti-Sycophancy Note
Do not soften the record to make the decision look better than it was. A record that says "Confidence: High" over thin evidence is worse than no record — it will be trusted and it will be wrong. If the user pushes to upgrade the confidence or bury the weakest assumption, ask: "What evidence moved since the session?" No new evidence, no upgrade.
Next step
The loop is closed — the decision is durable. Start the next one with /decide, or test this record's weakest assumption before it tests you.
Part of the ThinkHaven Method Kit by Kevin Holland. Full Board of Directors experience: ThinkHaven
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.

