verify-before-done
Skill2 starsChanged 8 months agoMITView on GitHub
Before claiming completion: check the whole agreed outcome, verify the result with appropriate evidence, and preserve unfinished work.
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
- MPIsaac-Per/claude-code-loop-patterns★ 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 repository15 other files it gives its agents.
- AGENTS.mdAGENTS.md
- examples/CLAUDE.mdCLAUDE.md
- claudeceptionSkill
- closeSkill
- ideaSkill
- investigateSkill
- office-hoursSkill
- plan-ceo-reviewSkill
- reflectSkill
- retroSkill
- reviewSkill
- scopeSkill
- shipSkill
- startSkill
- tddSkill
The file
We split this skill into the rules it gives an agent
5 rules. 0 are common, 5 are distinctive, and 0 are very specific to this project.
0
Common
Most of the 10 edited versions we read keep these.
5
Distinctive
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 15
Before saying a task is complete, handing it to someone else, saving/uploading a release, or marking work done.
- Distinctive: line 20
Recover the scope: Read the original request, accepted changes and current brief.
- Distinctive: line 23
Choose evidence that fits each outcome: Use the table below. A test command is one kind of evidence, not a requirement to invent tests for …
- Distinctive: line 25
Perform the check and read the result: Use current evidence for the exact artifact or target.
- Distinctive: line 69
Do not say work is continuing in the background unless its running or queued state was checked.
“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
Check the agreed outcome, then verify the result. A finished draft, passed test, review, or saved file proves only its own scope.
When to use
Before saying a task is complete, handing it to someone else, saving/uploading a release, or marking work done. Apply this to documents and operations as well as code.
The process
- Recover the scope. Read the original request, accepted changes and current brief. List the agreed outcomes. Do not silently drop an obligation because one phase finished. An optional idea is not an obligation; a canceled task must not be revived.
- Choose evidence that fits each outcome. Use the table below. A test command is one kind of evidence, not a requirement to invent tests for every document.
- Perform the check and read the result. Use current evidence for the exact artifact or target. If reusing an earlier check, confirm the artifact and relevant state are unchanged. Missing tools, unavailable evidence, failed commands and empty test collection are not passes.
- Reconcile what remains. Mark each agreed outcome complete with evidence, ready for a named next action, waiting for a specific dependency, or removed/deferred by the user or an already-agreed scope condition. Preserve unfinished work in the existing brief/checkpoint.
- Act or hand back honestly. Continue authorized, executable work. If there is a real blocker, user pause or tool limit, state the next owner/action or resume trigger. Respect existing permissions and other writers; this skill grants no new access or authority.
Evidence by outcome
| Claim | Appropriate evidence | Not enough |
| A document meets the request | Compare every requirement with the text; verify source facts, calculations and cited links as applicable | The document exists |
| A table, chart, slide or PDF is usable | Inspect the rendered output and its labels; check underlying values and provide a text equivalent when useful | Its source file parses |
| A file is saved | Read the exact destination back and confirm the intended contents | Text prepared in chat |
| Work can resume in another tool | The new tool can recover decisions, exclusions, current artifact and remaining work from the supplied sources/checkpoint | A shared folder or handoff was created |
| A bug is fixed | Reproduce the original problem and verify the changed behavior | A nearby test passes |
| Tests or a build pass | Run the applicable command; read its exit status, failures and collected results | “Should pass,” an old run, or suppressed errors |
| A record was updated, published or sent | After an authorized action, read back the exact target in the correct account and confirm content/state | An API acknowledgment or reviewer's approval |
| The whole task is complete | Every accepted outcome is verified or explicitly removed/deferred under the agreed scope | One completed phase, a green suite or a closed session |
For tests, keep failures visible. Do not substitute another runner merely because the selected command failed. If the project has no applicable tests, say so and use appropriate checks; do not call that “tests passed.” For consequential work, the kit's independent review still applies.
A useful handoff
Keep the checkpoint
in the existing brief. At an authorized wrap-up, the designated closer records the follow-up
in the existing memory/open.md row, or creates one if this session originated the
follow-up and no row exists, under the close rules.
Helpers return findings and never write shared memory.
Say what actually happened:
- “The draft is saved and checked. Review is still needed; nothing has been sent.”
- “The report is complete. The user explicitly deferred its translation.”
- “I prepared the text but could not write the file. It is not saved.”
- “The update was accepted by the API, but target read-back failed. Publication is unverified.”
A review does not authorize sending or publishing. A session ending does not finish its task. Do not say work is continuing in the background unless its running or queued state was checked.
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.