- Home
- Skills
- Agents & Orchestration
- session-handoff
Works with the AI Agents you already use
Session Handoff 1
Write a copy-pasteable handoff so unfinished work continues in a fresh session instead of an expensive one.
Free
session-handoff
Example session with this skill installed
Wrap this up. I'm out of time today and want to continue in a fresh session tomorrow without paying to re-read this whole thing.
- Read your context and instructions
- Compiled the session-handoff
Handoff written to .claude/handoff.md. Start tomorrow's session with: read .claude/handoff.md
Handoff — Move payment flow onto RabbitMQ events
State
Publisher side works: PaymentService emits payment.captured and tests pass. Consumer is half-done — listener wired, but the idempotency check isn't in yet, so a redelivered message double-books. DLQ untouched.
Files that matter
src/payment/PaymentService.java:88— publishespayment.capturedafter capture. Done.src/payment/CaptureListener.java:34— consumes the event. Missing: dedupe onpayment_idbefore processing.src/config/RabbitConfig.java:19— queues declared. DLQ bean still a TODO.
Decisions already made
- Events over a synchronous call — a slow downstream no longer holds the request open. Do not re-litigate.
- Dedupe by
payment_idin a processed-ids table, not by message id. Message ids change on redelivery.
Next
- Add the processed-id guard at
CaptureListener.java:34, reject duplicates before booking. - Declare the DLQ in
RabbitConfigand route failed consumes to it. - Add a redelivery test that asserts a single booking.
Gotchas
- Local broker advertises on 127.0.0.1; inside the container that's the container. Use the host alias.
Connects securely to your tools. The creator never sees your data.
About this skill
A long agent session gets expensive because every turn re-reads the whole thing. The fix is starting fresh — but only if the state survives the jump. This skill writes that bridge.
It produces a tight .claude/handoff.md: the goal, what works and what's half-done, the files that matter with line numbers, the decisions already made and why (so the next session doesn't re-derive them), the concrete next actions, and the gotchas that wasted time. Written for someone with zero memory of the conversation — names and paths, not "as discussed."
Use it when a session has grown too big to keep working in, when you want to /clear but keep the context, or when you're picking the work back up tomorrow. If nothing's outstanding, it tells you the work is done instead of writing a file.
No dependencies, no setup. One SKILL.md.
Handoffs are limited to ~80 lines to remain token-efficient. Does not generate a handoff if all tasks are completed. Requires filesystem write access.
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 1 month ago
- Free with an Agensi account
Permissions
File Scopes
Writes a single handoff file at .claude/handoff.md and, if not already ignored, appends that path to .gitignore. No network, no secrets, no reads outside the project.