handoff-rules
Skill2 starsChanged 25 days agoMITView on GitHub
Writes a HANDOFF.md when one agent stops and the next one takes over. The file records what was done, what is still open, what was tested and how, and who holds the task now. The next agent reads it before it acts. Fixes "the new agent started over, redid finished work and trusted a test nobody ran." Use when a session ends with work still left, before /clear, when you pass a task to a subagent or to another agent, and when you pick up work someone else started. Also on /handoff-rules.
In this repository102 other files it gives its agents.
- access-levelsSkill
- agent-failuresSkill
- agent-guardSkill
- agent-hooksSkill
- agent-limitsSkill
- agent-loopSkill
- agent-patternsSkill
- agent-wikiSkill
- ai-in-businessSkill
- answer-checksSkill
- ask-firstSkill
- automate-questionsSkill
- batch-judgeSkill
- big-archiveSkill
- brief-partsSkill
- bug-triageSkill
- build-an-evalSkill
- calls-to-keepSkill
- checkpointsSkill
- check-waysSkill
- claude-md-rulesSkill
- code-mapSkill
- command-guardSkill
- compact-safelySkill
- context-rulesSkill
- corrected-twiceSkill
- count-checkSkill
- data-leaksSkill
- day-gateSkill
- disprove-findingsSkill
- dogfood-firstSkill
- done-testSkill
- draft-only-mailSkill
- drift-checkSkill
- fail-closedSkill
- failed-readsSkill
- faster-sameSkill
- first-messageSkill
- fix-reviewSkill
- handoff-noteSkill
- hard-problemsSkill
- harness-starterSkill
- hidden-text-checkSkill
- hooks-to-addSkill
- injection-guardsSkill
- inventory-firstSkill
- keys-and-agentsSkill
- live-checkSkill
- lock-on-failureSkill
- long-videosSkill
- mcp-partsSkill
- memory-queueSkill
- memory-sourcesSkill
- memory-trimSkill
- night-alarmSkill
- objectorSkill
- old-code-judgeSkill
- one-task-gateSkill
- past-green-testsSkill
- phone-rulesSkill
- The first 60 shown.
The file
We split this skill into the rules it gives an agent
8 rules. About 0 are common, 5 are distinctive, and 3 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈5
Distinctive · estimate
This author's own choices. Worth a look.
3
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 13
If it is not in the handoff file, the next agent does not know it.
- Distinctive: line 14
Write down what is open and what was not tested.
- Very specific: line 19
Before you stop, write HANDOFF.md in the project folder.
- Distinctive: line 28
Write the handoff before /clear.
- Distinctive: line 29
Keep the file short enough to read in a minute.
- Very specific: line 30
If the owner: line is unclear, stop and ask the person.
- Very specific: line 31
Before you build on a pass in tested:, run that check again yourself.
- Distinctive: line 32
If the file and the code disagree, trust the code.
≈ “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
Handoff rules
The new agent started over. It redid finished work, missed what was left and trusted a test nobody ran. The first agent knew all of it, but none of it was written down. This skill writes one handoff file when an agent stops, and the next agent reads that file before it does anything.
The principles
- If it is not in the handoff file, the next agent does not know it.
- Write down what is open and what was not tested. A gap in the file reads as done.
- One agent holds the task at a time. Two holders make two versions.
- The next agent reads first and acts after.
Steps
- Before you stop, write HANDOFF.md in the project folder. Put the task in one sentence at the top, and under it an
owner:line that names who holds the task now. - Under
done:, write what was done: each change, the files it touched and each decision with its reason. Write only facts the next agent can check. - Under
open:, list the work that is still open, in the order it should be done, and say what blocks each item. If you tried something and dropped it, say what and why, so nobody tries it again. - Under
tested:, list what was tested and how: the command or the steps, and the result. Then list what was not tested, and why. Anything not run is marked not run, never passed. - Pass the task. Change the
owner:line to the next agent, then stop editing. The old session does not touch the files again. - The next agent reads HANDOFF.md first and runs
git statusandgit logto check it. In two lines it says what it understood and what it will do first. Then it acts. - Each agent that stops updates
done:,open:,tested:andowner:.
Rules
- Write the handoff before /clear. Write it before /compact too, so it holds what the summary might drop.
- Keep the file short enough to read in a minute. Point to logs and files instead of pasting them in.
- If the
owner:line is unclear, stop and ask the person. Two agents never work on the task at once. - Before you build on a pass in
tested:, run that check again yourself. - If the file and the code disagree, trust the code. Fix the file and say what you changed.
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.

