flow-handoff
Skill1 starsChanged 48 days agoMITView on GitHub
Resumable handoff record for another session. Use when pausing or ending a session mid-task, or saving an interrupted plan or run. Routes to `flow-orchestrate`, `flow-work` or the named specialist.
In this repository38 other files it gives its agents.
- AGENTS.mdAGENTS.md
- llms.txtllms.txt
- flow-a11ySkill
- flow-adversarial-reviewSkill
- flow-animateSkill
- flow-apiSkill
- flow-benchmarkSkill
- flow-brainstormSkill
- flow-commitSkill
- flow-data-migrateSkill
- flow-depsSkill
- flow-designSkill
- flow-diagnoseSkill
- flow-docsSkill
- flow-git-fixSkill
- flow-migrateSkill
- flow-observeSkill
- flow-onboardSkill
- flow-orchestrateSkill
- flow-planSkill
- flow-pr-fixSkill
- flow-prototypeSkill
- flow-prune-agent-setupSkill
- flow-redesignSkill
- flow-refactorSkill
- flow-releaseSkill
- flow-repo-briefSkill
- flow-reviewSkill
- flow-shipSkill
- flow-simplifySkill
- flow-specializeSkill
- flow-stylesSkill
- flow-testSkill
- flow-triageSkill
- flow-uiSkill
- flow-verifySkill
- flow-workSkill
- flow-write-skillSkill
The file
We split this skill into the rules it gives an agent
1 rule. About 0 are common, 0 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈0
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
- Very specific: line 17
Write to the user's location or .handoffs/<date>-<topic>.md, preserving ignore policy.
≈ “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
Give a fresh agent a reliable starting point: current intent, evidence and unresolved work, not a retelling of the conversation.
Process
- Identify the user's latest goal and steering, branch/revision, local changes, live plan and any orchestration run. Do not replace the goal with the last status question.
- Write to the user's location or
.handoffs/<date>-<topic>.md, preserving ignore policy. A Git-ignored file stays on this machine; share it or use an agreed tracked location when the next session runs elsewhere. A path alone does not transfer a file. - Record decisions with rationale, source paths, accepted and pending work, dead-ends, missing prerequisites and the exact next action. Reference plan and run records instead of duplicating them.
- Include the verification revision and evidence location, not a list of green commands. For orchestration, keep worker/verifier identities, rejected attempts, budget consumed, stale acceptance and in-flight actions. The next session must not reset limits.
- Name the resume command and first inspection: compare current source with recorded inputs, reconcile in-flight operations, and inspect unknown postconditions before replaying non-idempotent actions.
Handoff record
# Handoff - <goal>
Status: in-progress | blocked | review-pending | complete
Repository: <branch, commit, relevant staged/unstaged/untracked changes>
Current intent: <goal, non-goals, latest user constraints>
Plan/run: <paths, revision, direct or orchestrated>
Accepted work: <tasks, verifier/evidence, revision>
Pending/rejected work: <task, reason, next action>
In-flight actions: <identity, known outcome, how to inspect>
Budget: <attempts/time consumed, remaining limits>
Decisions and dead-ends: <why, not transcript>
Required access: <capability names, no credentials>
Resume: <skill + artifact path + first check>
Artifact availability: <local-only or shared location>Skip fields that do not apply, but never omit an unresolved dependency or unknown external outcome. The live plan carries the checklist; the handoff carries the context to trust it in the next session.
Anti-patterns
- Calling an ignored directory durable across machines because it sits inside a repo.
- Resuming at the next unchecked task while its dependency evidence is stale.
- Dumping the transcript or copying credentials.
- Using a new session to reset a failed correction loop.
Done when
- The file exists and its actual availability is stated.
- The next session can recover intent, accepted evidence, remaining work and blockers.
- The first resume action checks state before further changes.
Resume an existing run with flow-orchestrate, a direct plan with flow-work, or the documented
specialist for diagnosis, research or review.
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.

