evidence-discipline
Skill1 starsChanged 24 days agoMITView on GitHub
Use when work is about to be called done and the claim needs proof — "mark this step done", "CI passed, close it out", "how do I prove this is finished", "just use the last green run as evidence", "what's left before we ship" — and before YOU say a task is complete. A step closes only on a real, green CI run of this code (never a placeholder URL, pasted command text, or an older green run), reversals are recorded rather than erased, and the 27-step conductor ledger (/conductor) enforces it.
3 other projects use a version of this.
Adopted by
Most recently taken up by
- curvedinf/dev-skillstheir own version, first seen Oct 20, 2026
- rikvanriel/code-hygienetheir own version, first seen Sep 17, 2026
- tatargabor/set-coretheir own version, first seen Mar 19, 2026
Other versions3 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 3
- tatargabor/set-core★ 36same name
- curvedinf/dev-skills★ 7same name
- rikvanriel/code-hygiene★ 2same 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 repositoryOne other file it gives its agents.
- eval-gatesSkill
The file
We split this skill into the rules it gives an agent
4 rules. About 0 are common, 4 are distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈4
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 8
Use this skill when the user works with the conductor ledger, asks how to prove progress, or when YOU are about to claim a step of work is …
- Distinctive: line 26
Never construct the URL from memory, never paste the command that WOULD produce evidence as if it were evidence.
- Distinctive: line 28
If the run is red, the step stays open.
- Distinctive: line 48
When you complete work tracked by a ledger: run the verification, capture the real evidence, and only then check the step.
≈ “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
Evidence discipline (the conductor method)
Use this skill when the user works with the conductor ledger, asks how to prove progress, or when YOU are about to claim a step of work is done.
Why it exists
The ledger's design assumption is that the most likely source of fake
evidence is the person (or model) doing the work, under schedule pressure —
not an attacker. During its own development this system closed steps on a
literal <PASTE-REAL-RUN-ID> placeholder and on pasted command text, twice,
before the evidence check was hardened. Every caught fake is preserved in
the committed state log as a caught-and-voided entry. The audit trail is
the feature. Never offer to clean, squash, or rewrite it.
Operating rules
- A step closes with
check --evidence <proof>. For CI-kind steps, proof is a realhttps://github.com/<org>/<repo>/actions/runs/<id>URL. Obtain it programmatically (Actions API or page) AFTER the run finishes green. Never construct the URL from memory, never paste the command that WOULD produce evidence as if it were evidence. - If the run is red, the step stays open. Report the failure; don't shop for an older green run of different code — evidence must match the work it claims to prove.
- Deferrals need a written trigger ("fires when ..."); N/A needs an architectural justification. T05 (instrument governance) and T27 (audit the eval system itself) can never be N/A.
- Mistakes are handled with
reopen, which voids the evidence and logs the reversal. That log entry is not embarrassing; it is the point. auditruns in CI, release-blocking. A ledger that only the author reads is a diary, not a control.
Honest limit
The ledger verifies evidence SHAPE (a real Actions-run URL), not ownership — it cannot prove offline that a URL belongs to this repo's run of this code. That last step belongs to the human reviewer reading the linked run. Never present shape-validation as authenticity.
For Claude specifically
When you complete work tracked by a ledger: run the verification, capture the real evidence, and only then check the step. If you cannot obtain real evidence (network down, CI not finished), say exactly that and leave the step open — an honest open step costs a little time; a fake closed one costs the whole ledger's credibility. These rules apply to any evidence-tracked workflow, not only this plugin's files.
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.