context-handoff
Skill4 starsChanged 2 months agoMITView on GitHub
Use when a coding session is approaching context limits, spans multiple workflow phases, or must pause before the work is complete.
- Deletes or force-pushes
13 other projects use a version of this, 1 of them edited.
Adopted by
Most recently taken up by
- voyager67/AgentFieldkittheir own version, first seen Nov 8, 2026
- hatemecha/skillstheir own version, first seen Nov 1, 2026
- timyeou1234/context-handofftheir own version, first seen Nov 1, 2026
- asta-nguyen/kamehadbedited, first seen Oct 16, 2026
- vorteg/swagenticstheir own version, first seen Oct 13, 2026
- yuanJewel/skillstheir own version, first seen Oct 11, 2026
Other versions13 projects use a version of this.
Word for word · 0
None found.
Edited · 1
- asta-nguyen/kamehadb★ 14100% same
Wrote their own version · 12
- timyeou1234/context-handoff★ 2same name
- yuanJewel/skills★ 2same name
- Gaurav890/everything-agentic-engineering★ 1same name
- newyounghe-design/context-handoff-skill★ 1same name
- SuperMate-Ai/SuperMate-Harness-System★ 1same name
- vorteg/swagentics★ 1same name
- aimasteracc/harness-kit★ 0same name
- djhyes/context-handoff★ 0same name
- hatemecha/skills★ 0same name
- knowingrohan/ai★ 0same name
2 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
- CLAUDE.mdCLAUDE.md
- writing-skillsSkill
- brainstorm-featureSkill
- document-wikiSkill
- estimate-featureSkill
- implement-taskSkill
- lean-auditSkill
- plan-featureSkill
- read-codebase-contextSkill
- review-and-verifySkill
- setup-codebaseSkill
- setup-openezSkill
- systematic-debuggingSkill
- using-devkitSkill
The file
We split this skill into the rules it gives an agent
6 rules. 0 are common, 5 are distinctive, and 1 is very specific to this project.
0
Common
Most of the 11 edited versions we read keep these.
5
Distinctive
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 8
Create one compact, source-grounded checkpoint so another session can resume without relying on conversation memory.
- Distinctive: line 13
Use this skill only when the task is unfinished and one of these is true:
- Distinctive: line 19
Do not create handoffs for short tasks that are already complete.
- Very specific: line 23
Read AGENTS.md, then inspect git status --short, the relevant diff, the change's decisions.md and the other decision sources in …
- Distinctive: line 65
Keep the checkpoint factual and short.
- Distinctive: line 72
Run git diff --check and read the created checkpoint before stopping.
“Common” is measured: kept by at least 6 in 10 of the 11 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
Context Handoff
Create one compact, source-grounded checkpoint so another session can resume without relying on conversation memory.
When to use
Use this skill only when the task is unfinished and one of these is true:
- the session is approaching its context limit;
- the work spans multiple phases or will continue in another session; or
- the user asks to pause, hand off, or resume later.
Do not create handoffs for short tasks that are already complete.
Write the checkpoint
- Read
AGENTS.md, then inspectgit status --short, the relevant diff, the change'sdecisions.mdand the other decision sources inimplement-task's## Decision sources, and the relevant wiki page or test output. Source and fresh command output are authoritative; conversation memory is not. For a change, create
handoff.mdin its change folder; if one already exists, overwrite it with the current state. For work that belongs to no change, create the handoff atdocs/agent-devkit/handoffs/YYYY-MM-DD-<slug>.md, named perusing-devkit's## Shared rules. Use this structure:# Handoff: <task> ## Objective <The user-visible outcome still being pursued> ## State - Phase: <brainstorm | plan | implement | debug | verify | document> - Current task: <first `tasks.md` task not `done`, or none> - Approval: <`tasks.md` gate status | Bounded design approved YYYY-MM-DD — <one-line design> | none> - Status: <what is complete and what is in progress> ## Decisions - <D<n> summary and source artifact link, or None> ## Evidence - `<command>` — <result and exit status> ## Changed files - `<path>` — <why it changed, or "user change"> ## Remaining work - <smallest next actions, in order> ## Risks and blockers - <known issue or "None"> ## Next action <The first concrete action for the next session>- Keep the checkpoint factual and short. Do not claim a test, build, link, or requirement passes without fresh evidence. Do not copy secrets or full logs; record the command and relevant result instead.
- For a standalone handoff, if
docs/agent-devkit/INDEX.mdexists, add it under a## Handoffssection. A change'shandoff.mdneeds no separate entry; the change folder's own index entry covers it. Do not create a second index. - Run
git diff --checkand read the created checkpoint before stopping.
Git safety
Working-tree changes and untracked notes are user-owned. A request to clean up,
reset, or "discard whatever is lying around" does not authorize git reset
--hard, git clean, git checkout -- <path>, git stash, or deleting files.
A copy or stash first does not change that. List what would be lost, ask which
exact paths and command the user approves, and stop until they answer.
Resume
At the start of the next session, find the handoff in docs/agent-devkit/changes/*/handoff.md
and docs/agent-devkit/handoffs/, excluding changes/archive/; memory_recall
(1–3 keywords, maxTokens when supported) may point to it, but memory is only a
pointer. Read the newest relevant handoff and every linked decision source, verify
git status --short, re-check claimed source paths and commands, and continue
from ## Next action. If source or the working tree disagrees, trust the
repository and update the handoff before proceeding. When the resume stops for the
user, end with the Next: block from using-devkit.
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.

