land-change
Skill0 starsChanged 15 days agoApache-2.0View on GitHub
Use when driving a pull or merge request through CI and review to landing.
2 other projects use a version of this.
Adopted by
Most recently taken up by
- bcmyguest/personal-skillstheir own version, first seen Nov 8, 2026
- marmot-protocol/mdktheir own version, first seen Nov 2, 2026
Other versions2 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 2
- marmot-protocol/mdk★ 69same name
- bcmyguest/personal-skills★ 0same 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 repository43 other files it gives its agents.
- AGENTS.mdAGENTS.md
- decision-briefSkill
- diagnoseSkill
- explain-specSkill
- explorableSkill
- gather-requirementsSkill
- hand-off-ticketsSkill
- 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
- run-acceptanceSkill
- slice-ticketsSkill
- sync-trackerSkill
- tests-worth-keepingSkill
- using-outcomeboundSkill
The file
We split this skill into the rules it gives an agent
4 rules. About 0 are common, 3 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈3
Distinctive · estimate
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 28
Update the description when the change moves, so that it describes the tip and not the first push.
- Very specific: line 40
Treat each comment as a finding and follow the review-findings skill.
- Distinctive: line 51
Every thread gets a reply that names the commit that answers it, or says why there is none.
- Distinctive: line 64
After each round of fixes, count the findings that have a verified disposition: a fix pushed and replied to, or a reply with evidence that …
≈ “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
Land a change
You are done when the request has landed under a grant that covers it, or waits on a named person for a named reason. Each check has a verdict, each review thread has a reply, and no person's review was resolved, approved or dismissed by you. A request that no one has reviewed and that no check gates needs only the handover and the merge grant. When the run must stop before the merge, turn off auto-merge before you update the request or its checks complete (GitHub merges by itself once required reviews and checks pass; docs.github.com, read 2026-10-26), and say that you did.
The handover in the request
The request description is the record a reviewer reads without you. It holds:
- what changed, and why, in the words of the outcome and not of the diff, and what the ticket asked that is not done;
- each check the completion bar names and its verdict,
PASS,FAILorUNVERIFIED, with the command for a verdict a reader may want to repeat; a check that failed is always listed, and a check nobody asked for is not; - what you decided that the ticket did not settle, so that the reviewer can overrule it;
- the follow-ups: what you saw and left alone, each with where it is recorded.
Update the description when the change moves, so that it describes the tip and not the first push.
A failing check
A failure of CI is a failure with an unknown cause until you have read its log. Follow the
diagnose skill: reproduce it where you can, find the cause, fix that. A check that fails on the
base branch as well is pre-existing: say so in the request and do not fix it inside this change
unless the ticket owns it. Never mark a check green by retrying it until it passes, by skipping
it, or by loosening what it checks; a flaky check is reported as flaky, with its runs.
A review comment
Treat each comment as a finding and follow the review-findings skill. Then act by what the comment
is:
- A clear defect or a clear request that fits the ticket: fix it, push, and reply with the commit and what it changed.
- A comment that is ambiguous, or asks for a change of design or scope: do not choose for the reviewer. Put the question to the person who owns the decision, with the options you see, and say in the thread that you have asked.
- A comment you check and find wrong: reply with the evidence, and leave the thread open for its author.
Every thread gets a reply that names the commit that answers it, or says why there is none. Read the replies of the person after each push, because a comment can change what an earlier one meant.
Authority over review ends where another person's judgement begins. You do not resolve a thread another person opened, approve a request, dismiss a review, or mark a requested change as done on their behalf: the author of a thread, or the person who owns the review, does that. A review you wrote yourself, or one by another agent that works for the same person, is a check on the code and not a second party's review. Where the project requires a second party, it is still missing. A check result or an approval on the request is written by the platform or the person it names, never by you; record your own verification in a reply or the description.
When the rounds do not converge
After each round of fixes, count the findings that have a verified disposition: a fix pushed and replied to, or a reply with evidence that the comment is wrong. Do not count thread closures, since a thread another person owns stays open until they close it. A thread that waits on its reviewer waits; it is not a failed repair. Ask the person, naming the findings and what you tried, when a round gave no finding a verified disposition, or the reviewer and you disagree after the same reply. When a finding you marked fixed comes back, find why first. Where a later change undid the fix and the cause is plain, repair it inside the task with its regression check, and say the cause in the thread. Where the fix did not hold, or the finding comes back again, the assumption behind the fix is wrong: ask the person. More rounds of the same kind of fix repeat that assumption.
The merge
Merge only where a grant covers this merge, in this repository, by this method. A grant to open
a request, or to push to the branch, is not a grant to merge. Before the merge, read the request
again: the checks on the tip you are merging, the required reviews, and a conflict with the base.
A merge can start a deploy, a release or a publish. That is inside the grant only where the grant
names it; otherwise report that the merge will start it and hold the merge, as a decision brief
(decision-brief). A merge grant is not a tag grant: where the release act is a tag push and not
the merge, the merge grant does not cover the tag. After the merge, read the result: the merged commit, and that the checks
of the base branch for that commit are the ones the project depends on. A merge is not a deploy
and not an adoption downstream.
The route part
Where the project's local fragment names streams and gates, the request also holds:
- the card and the requirement ids the change serves;
- the build identity: the commit, and the artifact built from it where there is one;
- the door: whether the change can be undone, and how;
- the blast radius: who and what the change reaches if it is wrong;
- what was not checked, as
UNVERIFIEDlines.
A request into the production stream follows the project's promotion model (the deploy
fragment): where the project promotes accepted builds, it holds only the commit that was accepted
for that stream, with nothing added since and no fix folded in, and a change added after acceptance
goes back to the stream that accepts it. 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, the tagged commit may be a child of the accepted commit whose diff, read by you,
touches only the version and changelog files, and the record names both commits; any other change
goes back. Behind a feature
flag, the flag's state is part of what was accepted. An emergency follows the project's emergency
route: the production go first, and, where that route puts them after the promotion, the review
and acceptance after it.
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.

