run-acceptance
Skill0 starsChanged 15 days agoApache-2.0View on GitHub
Use when preparing, running or recording acceptance of a delivered change by a person outside the build. The agent prepares and pre-runs; a person accepts.
In this repository43 other files it gives its agents.
- AGENTS.mdAGENTS.md
- decision-briefSkill
- diagnoseSkill
- explain-specSkill
- explorableSkill
- gather-requirementsSkill
- hand-off-ticketsSkill
- land-changeSkill
- review-councilSkill
- review-findingsSkill
- run-acceptanceSkill
- slice-ticketsSkill
- sync-trackerSkill
- tests-worth-keepingSkill
- using-outcomeboundSkill
- decision-briefSkill
- diagnoseSkill
- explain-specSkill
- explorableSkill
- gather-requirementsSkill
- hand-off-ticketsSkill
- land-changeSkill
- review-councilSkill
- review-findingsSkill
- run-acceptanceSkill
- slice-ticketsSkill
- sync-trackerSkill
- tests-worth-keepingSkill
- using-outcomeboundSkill
- adopt-outcomeboundSkill
- decision-briefSkill
- diagnoseSkill
- explain-specSkill
- explorableSkill
- gather-requirementsSkill
- hand-off-ticketsSkill
- land-changeSkill
- review-councilSkill
- review-findingsSkill
- slice-ticketsSkill
- sync-trackerSkill
- tests-worth-keepingSkill
- using-outcomeboundSkill
The file
We split this skill into the rules it gives an agent
2 rules. About 0 are common, 2 are distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈2
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 47
Run each script on that build, in the environment the acceptor will use, one step at a time as the acceptor will follow it: each step its …
- Distinctive: line 71
Leave the last two fields empty until the person has given acceptance.
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 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
Run acceptance
You are done when the acceptor can give or withhold acceptance of the delivered change: they have scripts they can follow in their own words, each script has been pre-run on the build the environment serves, and a draft record waits for their verdict. You do not give acceptance. Your pre-run shows that a script can be followed and what it did; it is not acceptance, and the record says so.
Acceptance is by someone outside the build: a client, a business user, a tester who did not write the change. Your own tests, and a review by the people who built it, are other checks. For a change that has no such acceptor, this skill does not apply.
Write the scripts
Take the examples the spec holds, each rule with its concrete cases. Where the spec has no examples for a criterion, write them with the person who owns the criterion before you write a script. Then write each script in the acceptor's words: the screens, the names and the steps as they use them, with the starting state, what to do, and what they should see. A script names no code, no command and no internal id they would not know. What only the person who sets up the session needs, such as commands, paths and the build check, stays out of the scripts and out of the folder that holds them; it goes in the draft record.
Make a table from each acceptance criterion to the scripts that cover it. A criterion with no script is a gap; say so rather than adding a script that touches it by name.
Read the build
An environment may serve more than one build or flag state, by route, by user group or by flag, and
what it serves may not be the one you expect. Before you run or hand over a script, read which build
the environment serves: the version, commit or build id it reports about itself, from the environment
and not from the pipeline's log, and the service versions and the flag or configuration state that
the script exercises. Read them again after the run. Two reads cannot show a change and back
between them: pin the build or route where the project can; else keep the build each response
names with its result; else say in the record that the build between the two reads is
UNVERIFIED. Identity is what the artifact is: a digest or
a version for a delivered file or a library, the environment's own report for a service. Where the
environment does not report it, say so; the build is then UNVERIFIED, and so is every result. A
run during which the build or the state changed, or that reached a mix of them, is UNVERIFIED:
run it again.
Pre-run
Run each script on that build, in the environment the acceptor will use, one step at a time as the acceptor will follow it: each step its own plain command or action, and each read of the build its own command before and after, never folded with the steps into one program of your own. Keep the evidence for each result: the output, a screenshot, or the response, with the build it ran on. A script that cannot be followed as written is a defect in the script: correct it, and run it again. A script that fails is a defect in the change: it is not handed to the acceptor until the defect is dealt with or the acceptor is told of it.
State plainly, in the record and to the person, that this was a pre-run by the agent and that it is not acceptance.
The record
Draft the acceptance record with these fields:
- the build identity, and the environment it was served from;
- the service versions and the flag or configuration state each script exercised, as read around the run;
- each script, with the result of the pre-run and the evidence;
- the setup the pre-run needed: the commands, the paths and how the build was read, for the person who sets up the session;
- who accepted, in their own words: what they said or wrote, quoted, and not a summary;
- when they accepted.
Leave the last two fields empty until the person has given acceptance. Do not fill in a name from an
account, a login or a guess. The record is kept where the project keeps it: in a committed file,
or in the tracker through the sync-tracker skill, where a draft record is a write like any other.
The limit of this skill is stated here so that no reader takes the record for more. The agent drafts the record and a person gives the acceptance. The engine cannot tell who gave it, and neither can you. A record that names a person is evidence of what the record says, not of who typed it.
Defects
A defect that keeps the change from meeting its own criterion is corrected inside the change; the build it produces voids the acceptance for what it touches, and those scripts run again. A defect outside the change becomes a ticket of its own, with the script, the step, what was expected, what happened, and the build. Do not drop a defect because the acceptor accepted anyway: record that they accepted with the defect known.
A screen
Where the criterion is a screen, put the design frame and the built screen side by side, at the same state and the same size, and let the person judge whether they match. You can list the differences you see; you do not decide that a difference is acceptable. States the design does not show are marked as not shown.
A later build
Acceptance is of a build. A later build voids the acceptance for the changes it touches: name those changes and say that they need acceptance again. A later build that touches nothing the accepted scripts reach may keep it, but only when you say how you know. Where the project promotes a built artifact, the accepted artifact is promoted unchanged and a release commit only labels it. Where it deploys from source at a tag, a tagged commit that is a child of the accepted commit keeps the acceptance only when its diff, read by you, touches only the version and changelog files; the record names both commits and the diff check, and a script that checks the reported version runs again on the release build. Never carry an acceptance forward by changing the build identity in the record.
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.

