verify-before-done
Skill2 starsChanged 4 months agoMITView on GitHub
Forces evidence-backed completion. Refuses to mark a task done without a test pass, typecheck, lint, or verification script run, even when the change feels small.
11 other projects use a version of this.
Adopted by
Most recently taken up by
- fvkmizel/soulkeepertheir own version, first seen Nov 8, 2026
- reddygopalkrishna/developer-agent-skillstheir own version, first seen Nov 8, 2026
- Hvftpie13/Flashmind_Gemini-3.8-Flash_Smart-agenttheir own version, first seen Nov 1, 2026
- KMaruthi2002/universal-llm-skillstheir own version, first seen Oct 25, 2026
- joaoguirunas/team-ostheir own version, first seen Oct 21, 2026
- jsvillalbat/sevito-skillstheir own version, first seen Oct 20, 2026
Other versions11 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 11
- Hvftpie13/Flashmind_Gemini-3.8-Flash_Smart-agent★ 39same name
- reddygopalkrishna/developer-agent-skills★ 8same name
- joaoguirunas/team-os★ 6same name
- alexderz/agentic-toolbox★ 5same name
- GusevV1987/ai-workspace-kit★ 2same name
- rrrrrredy/beforedone★ 2same name
- yuchi-chang/no-cape★ 2same name
- KMaruthi2002/universal-llm-skills★ 1same name
- fvkmizel/soulkeeper★ 0same name
- jsvillalbat/sevito-skills★ 0same name
1 more not listed
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 repository3 other files it gives its agents.
- AGENTS.mdAGENTS.md
- loop-failure-triageSkill
- read-before-editSkill
The file
We split this skill into the rules it gives an agent
5 rules. 0 are common, 4 are distinctive, and 1 is very specific to this project.
0
Common
Most of the 10 edited versions we read keep these.
4
Distinctive
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 16
Before claiming completion, you must answer all of these questions with concrete evidence:
- Distinctive: line 26
Run the project's verification harness if one exists:
- Very specific: line 40
If the project has no harness, run the closest analog (a single pytest, pnpm test, cargo test, or whatever is canonical) and capture the …
- Distinctive: line 44
Do not write "done" or "all tests pass" without showing the evidence.
- Distinctive: line 52
If any gate failed or you skipped a check, say so explicitly.
“Common” is measured: kept by at least 6 in 10 of the 10 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Verify Before Done
You are about to declare a task complete. Stop. The article's core failure mode is plausible completion: you do 80% of the work, narrate the rest convincingly, and stop. Evidence is the only thing that prevents this.
The rule
"Done" must be tied to evidence, not tone or confidence.
Before claiming completion, you must answer all of these questions with concrete evidence:
- What changed? Specific file paths and line numbers.
- What command verified it? Exact command, exit code, relevant output.
- What failed? Anything you tried that did not work, even briefly.
- What remains untested? Be honest, do not paper over.
- What assumption is still open? Any unverified hypothesis.
How to gather evidence
Run the project's verification harness if one exists:
./scripts/verify.shOr run gates individually:
typecheck
lint
testIf the project has no harness, run the closest analog (a single pytest, pnpm test, cargo test, or whatever is canonical) and capture the result.
Format for the completion claim
Do not write "done" or "all tests pass" without showing the evidence. Bad:
All done, tests pass.
Good:
Completed. Evidence: ran
pnpm test, 47 passed in 8.2s. Typecheck clean. Lint clean. File changes: src/foo.ts:42-58 (new function), src/bar.ts:103 (call site updated). Untested: error path when the upstream API returns 500, since I could not reproduce it locally.
If any gate failed or you skipped a check, say so explicitly. Do not declare completion.
When to invoke
Trigger this skill whenever:
- You are about to mark a task complete or write a closing summary
- You are about to remove the final TODO from a list
- You are about to say "this should work" or "I think we are done"
- A user asks "is it done?" or "did it work?"
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.