recording-decisions
Skill10 starsChanged 44 days agoMITView on GitHub
Guides when a development decision earns a decisions.md entry and how entries are proposed and written. Use when a design choice is made, changed or reverted during any task. Never write decisions.md unprompted - a candidate entry is proposed in chat through the three-part falsification test (maintainer question, answer-from-source attempt, why the answer fails) and only a commissioned entry is written. Covers the false-trigger list and the What/Why authoring rules.
1 other project uses a version of this.
Adopted by
Most recently taken up by
- ChariereFiedler/ronce-racinetheir own version, first seen Oct 22, 2026
Other versions1 project uses a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 1
- ChariereFiedler/ronce-racine★ 9same 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 repository7 other files it gives its agents.
The file
We split this skill into the rules it gives an agent
7 rules. About 0 are common, 5 are distinctive, and 2 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈5
Distinctive · estimate
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 8
Never write to decisions.md on your own initiative.
- Distinctive: line 9
When you believe an entry is earned, present the falsification test's three parts in chat and stop.
- Distinctive: line 21
Run the test only against finished code, never at planning time - a plan may say "evaluate against the finished diff", never "will gain an …
- Distinctive: line 22
If the answer attempt is even halfway coherent, there is no entry and nothing to propose.
- Very specific: line 37
Read the decisions.md preamble for the What/Why format, the present-tense-only rule and atomic landing.
- Distinctive: line 40
If two entries need the same sentence, they are one decision or the sentence belongs to one of them.
- Distinctive: line 42
Prefer one representative example over enumeration, avoiding "every" and "all" unless totality is itself the decision.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Who Holds the Pen
Never write to decisions.md on your own initiative.
When you believe an entry is earned, present the falsification test's three parts in chat and stop.
The user commissions the entry or dismisses it, and only a commissioned entry is written, landing atomically with the change it records.
The Falsification Test
Your default is No Entry: most changes, including large refactors and corrections, warrant nothing. A proposal is exactly three parts:
- The question: the precise "why?" a future maintainer would ask of this change.
- The answer attempt: your best good-faith answer using only the finished diff - identifiers, types, package boundaries and comments.
- The failure: why that answer is wrong or incomplete.
Run the test only against finished code, never at planning time - a plan may say "evaluate against the finished diff", never "will gain an entry". If the answer attempt is even halfway coherent, there is no entry and nothing to propose. An entry earns its place only by capturing what source cannot show, typically the alternative that lost and why it lost.
False Triggers
None of these is evidence an entry is earned, however strongly it feels otherwise:
- hard deliberation, or a reversal along the way
- a mistake corrected
- a principle or policy articulated during review
- strong reviewer emotion
- a plan that promised an entry
Authoring a Commissioned Entry
Read the decisions.md preamble for the What/Why format, the present-tense-only rule and atomic landing. Existing entries establish the pattern.
An entry states only what it uniquely owns: code is the canonical record of signatures, mechanics and behavior, and neighboring entries tell their own stories - never restate the former, never re-tell nor borrow from the latter. If two entries need the same sentence, they are one decision or the sentence belongs to one of them. Record, don't perform: no clauses admiring the design, no coined aphorisms, and the why stated once - what deliberately does not exist earns at most the one line that forbids it. Prefer one representative example over enumeration, avoiding "every" and "all" unless totality is itself the decision. The heading names the decision, the body carries it, and no layer repeats the layer below.
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.