Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add commands/longthread/claude-plugins/handoffgit clone --depth 1 https://github.com/longthread/claude-pluginsWhat it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00017 | $0.01555 |
| Opus 5 | $0.00009 | $0.00777 |
| Sonnet 5 | $0.00003 | $0.00311 |
| Haiku 4.5 | $0.00002 | $0.00155 |
Grade A, and why
handoff scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured yesterday.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 120 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Close out the current thread of work on programme $1 (resolve as /programme:resume does).
A handoff whose structure is right and whose content is a changelog is worthless. What makes one valuable is what got noticed during the work. Ask accordingly.
1. Measure the state yourself
Read .claude/session-continuity.json if it exists for docsRoot and compareBranch, as
/programme:init does. Never accept a figure from this conversation — including one you stated
earlier.
git fetch
git status --short
git log --oneline -10
git rev-list --count "<compareBranch>..HEAD"
git rev-parse HEAD
Run every gate in the ledger's Gates table and record the real output in last run and result.
If the Gates table is empty, ask what this project's gates are and fill it, including the
trust column and the flake rate for any gate that has one.
2. Interview — do not fill blanks
Ask these one at a time. They are not interchangeable with "what did you do?", which produces a changelog.
- What did you measure? Not what you did — what number, output, or observation do you now have that you did not have before?
- What did you get wrong, and what would the next session repeat if you do not say so?
- Which claims already in the ledger are now false?
- What is the next terminal condition — a command, and the output that means it is done?
- Does the next phase still sit where
## The arcsays — and is any later phase's work cheaper taken now? Ask it against the table, row by row, not from memory. Two shapes to name out loud: work this phase would build that aplannedphase later deletes, and work scoped out as "a later phase's" that costs less inside this one than it will on its own. If the answer is "nothing to change", name which rows you read to conclude that — the failure this question exists to catch is answering it from memory.
On question 2: if the session changed anything — code, plan, or doc — and the answer is "nothing", ask again. Rephrase toward the concrete — a wrong assumption, a review finding, a test that passed while checking nothing, a plan defect. A session with no mistakes worth recording is usually a session that did not look; do not settle for an empty answer because it is easier to write.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- yesterday First seen · 120 lines · 17 tokens per session scan A 3629bf01cc38
handoff is a command published in the GitHub repository longthread/claude-plugins (1 stars, last pushed 11d ago), licensed MIT. It adds 17 tokens to every session and 1,555 once invoked, about $0.0001 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other commands, from other repositories
save
Save this conversation as a new or existing reusable context.
import
Import a shared context bundle, or take a teammate's newer copy of one you already have.
use
Connect a NeatContext context to this session.
export
Export a saved context as a bundle folder you can share.
mode
Choose how this session may switch contexts on its own.
extensions
Show what the connected context expects to reach, and whether this machine provides it.