- Home
- Skills
- Agents & Orchestration
- Claude State: Verified Session Handoff
Claude State: Verified Session Handoff
Save unfinished coding work and resume with verified progress, pending tasks, and next steps.
Free
Works with the AI Agents you already use
Claude State: Verified Session Handoff
Example session with this skill installed
Illustrative example using supplied project facts: create a handoff for /workspace/demo-store on branch fix-cart-total. The goal is to correct checkout totals; the latest correction is to keep displayed prices VAT-inclusive. src/cart/total.ts and tests/cart/total.test.ts were changed. npm test -- total reported 12 passing tests. Browser checkout has not been tested. Pending: verify the checkout screen and add a rounding regression case. No background jobs are running. Do not claim any other checks passed.
- Read your context and instructions
- Compiled the claude state
Illustrative handoff
This example shows the structure of a checkpoint using the facts supplied in the prompt. It is not a record of a live customer session.
Task and latest request
Correct checkout totals while keeping displayed prices VAT-inclusive.
Project state
- Project:
/workspace/demo-store(example path). - Branch:
fix-cart-total. - Commit: not supplied; verify it before resuming.
- Changed files:
src/cart/total.tsandtests/cart/total.test.ts.
Progress and evidence
The totals implementation and its unit tests have been edited. The supplied test result is npm test -- total: 12 passing tests. This does not establish that the browser checkout works.
Pending work
- Inspect the current diff and confirm that VAT-inclusive display behavior is preserved.
- Add a regression test for the rounding case.
- Run the relevant tests again and verify the checkout screen.
Blockers and running work
No blockers were supplied. No background jobs are running. Do not infer additional completed checks or failed approaches.
Resume steps
Confirm the project and branch, inspect changed files, and recheck whether this checkpoint still matches the latest request. Then continue with the rounding regression case. Record new results in .handoff/current.md, keeping unrun checks explicitly marked as unrun.
Connects securely to your tools. The creator never sees your data.
About this skill
Keep an unfinished coding task ready for the next session. Session Handoff helps an agent save a concise local Markdown checkpoint, then verify that checkpoint before continuing.
What it captures
- The latest user request, including corrections.
- Project/worktree location and available Git branch, commit, and changed files.
- Completed work with evidence, checks actually run, and their outcomes.
- Pending tasks, blockers, useful decisions, and the next concrete action.
- Identifiers for work still running, so a new session can check it without launching duplicates.
How to use it
Ask your agent to save a handoff before switching sessions. It uses an established project location or .handoff/current.md. In the next session, ask it to read the handoff, verify the current project and files, and continue the work you authorized.
Included
A portable SKILL.md, an MIT license, and a reference for the optional claude-state integration. The portable workflow needs an agent with local file tools. Git and a terminal help verify repository state.
Automatic capture and restore hooks require the separate Claude Code integration; installing this skill alone does not register hooks or intercept compaction. The reference explains that setup.
Limits
Handoffs are lossy and do not replace Git commits, backups, or durable project memory. The skill instructs agents to exclude credentials and private transcript dumps, and to distinguish attempted actions from verified results. Review a handoff before sharing it. Live behavior across different agents has not been benchmarked.
Free and open source under MIT.
A handoff is a lossy checkpoint, not a backup or full conversation restore. Saving must be requested unless the separate Claude Code hooks are installed. This ZIP does not install those hooks, change .gitignore, or automatically sanitize existing files. Agents are instructed to exclude secrets, but you should review handoffs before sharing. Live behavior across different agents has not been benchmarked.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
1 use
Times used to date
Free forever
No account required to browse
Trust & safety
Security scan
Security scan passed 29 days ago
- Free with an Agensi account
Requires
Permissions
File Scopes
Reads the selected project's files and writes a local handoff, normally .handoff/current.md. Terminal access is used when verifying Git state or running checks already authorized by the user. No connected account, network service, or API key is required for the portable workflow. Keep handoffs local and review them before sharing. Optional Claude Code hooks are installed separately.