compact-handoff
Skill15 starsChanged 32 days agoApache-2.0View on GitHub
Writes a handoff of the task's objective, scope, evidence and next action when a long conversation, compaction or handover to another session or agent could lose them, and resumes the task from it afterwards.
1 other project uses a version of this.
Adopted by
Most recently taken up by
- hishamashraf585/compact-handoff-skilltheir own version, first seen Aug 18, 2026
Other versions1 project uses a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 1
- hishamashraf585/compact-handoff-skill★ 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 repository10 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- adversarial-verificationSkill
- agent-security-scanSkill
- boundary-contractSkill
- merge-companionSkill
- necessity-gateSkill
- session-intent-analyzerSkill
- skill-evolutionSkill
- task-routerSkill
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 33
Before requesting compaction or transferring a task, the response model records the unresolved cause of a failing test, the rollback path …
- Distinctive: line 39
If the most recent test failed, it follows systematic-debugging for that failure.
- Distinctive: line 40
Before claiming completion or partial completion, it follows verification-before-completion.
- Very specific: line 44
All eight fields are required.
≈ “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
Compact Handoff
The response model uses this skill when a long conversation, a transfer to another agent or session compaction could lose the state of the task. This skill does not trigger compaction, install hooks or create background summaries.
Record the handoff
The response model fills the following block from the retained task state and the original evidence. When an authorized handoff file already exists, it writes the block there and invokes python3 <loaded-skill-directory>/scripts/compact_handoff.py <absolute-handoff-file> --json through the host's shell tool. Otherwise it keeps the block in the host's handoff or task context and sends that same text to python3 <loaded-skill-directory>/scripts/compact_handoff.py - --json on stdin; it does not create a report only to obtain a filename. The validator returns status, errors and warnings. On status=fail or exit 1, the response model corrects the named fields before submitting the handoff; a pass confirms only the block's structure.
[compact-handoff]
- objective: <current objective>
- completed: <completed work and supported scope>
- in-progress: <current item and unresolved conditions>
- next-step: <one next action>
- files-changed: <absolute paths>
- tests-run: <commands, results and original evidence source/time>
- forbidden-surface: <paths or actions not to touch>
- rollback: <nearest concrete rollback path>For an unresolved repair, the response model carries the failure, the affected acceptance criterion, the stage, the target, the cumulative failed rounds, the status of the reassessment after five failed rounds, the cause hypothesis and any stopping condition in in-progress or next-step. It keeps the original evidence with its source and time. It applies systematic-debugging's incident accounting to those values; the handoff does not create a new incident.
Pause compaction
Before requesting compaction or transferring a task, the response model records the unresolved cause of a failing test, the rollback path and the context the receiver needs, without secrets. If a required item is missing, the response model obtains it before the handoff.
Resume
- The response model that hands off the task passes the validated block through the host's context-handoff facility, or keeps it in the existing authorized task context when the host has no such facility. After compaction, the response model that resumes the task reads
objective,in-progressandnext-stepfrom that block. A delegate receives only the context its assigned subtask needs. - It compares the actual diff of the files in
files-changedwith the handoff. - If the most recent test failed, it follows systematic-debugging for that failure.
- Before claiming completion or partial completion, it follows verification-before-completion.
Validator and warnings
All eight fields are required. The validator rejects secret-like values without printing them. A tests-run value of not run or pending produces a warning: compaction may proceed, but completion is not established.
The response model must not invoke /compact automatically, create extra reports only for compaction or present the handoff as completion evidence. ../coding-convention/verification-before-completion/SKILL.md decides what proves completion.
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.

