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/thatsuperdev/llm-handoff/creategit clone --depth 1 https://github.com/thatsuperdev/llm-handoffWhat 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 | $0.00016 | $0.01415 |
| Opus 5 | $0.00008 | $0.00707 |
| Sonnet 5 | $0.00003 | $0.00283 |
| Haiku 4.5 | $0.00002 | $0.00142 |
Grade A, and why
create 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 — 222 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Create a complete handover.md in the project root so any LLM — Claude, Codex, Gemini, KiloCode, or any other tool — can continue this session without mistakes or repeated work.
Steps
1. Gather deterministic state
Run each of the following:
git rev-parse --show-toplevel
git branch --show-current
git log --oneline -10
git status --short
git diff --stat HEAD
git stash list
git worktree list
Also collect:
- Your own model name and version
- Which tool is running you (Claude Code, Codex CLI, etc.)
- Any running services:
lsof -i -P -n 2>/dev/null | grep LISTEN | head -20 - Relevant env var names (not values):
env | grep -E 'NODE_ENV|APP_ENV|DATABASE|API_URL|PORT' | sed 's/=.*/=<set>/' - Runtime versions:
node --version 2>/dev/null; python3 --version 2>/dev/null; go version 2>/dev/null; cargo --version 2>/dev/null
2. Write handover.md
Write to <git root>/handover.md using this exact structure:
---
generated: <ISO 8601 timestamp>
previous_llm: <your model name and version>
tool: Claude Code
trigger: manual
branch: <branch>
worktree: <path or N/A>
---
# LLM Handover
> Read this file completely before taking any action. Verify repository state matches
> what is documented before starting work. If anything conflicts, investigate first.
## Quick start checklist
- [ ] Read this entire file
- [ ] Run `git status` — confirm it matches **Repository state** below
- [ ] Run `git log --oneline -5` — confirm last commit matches
- [ ] Read the files listed in **Files in play**
- [ ] Start with **Next steps item 1**
---
## Original request
<verbatim or near-verbatim: what the user asked for when this session started>
## Session goal
<one sentence: what we are trying to achieve>
**Scope — in:**
- <what is in scope>
**Scope — out:**
- <what is explicitly out of scope>
**Constraints:**
- <constraints the next LLM must respect>
---
## Repository state at handoff
| Field | Value |
|-------|-------|
| Branch | `<branch>` |
| Worktree | `<path or N/A>` |
| Last commit | `<hash message>` |
| Working tree | <clean or list of dirty files> |
| Staged | <list or "none"> |
| Stashed | <list or "none"> |
**Uncommitted changes:**
<git status --short output>
**Recent commits:**
<git log --oneline -5 output>
---
## Progress
### Completed this session
- <specific task — file:line reference where relevant>
### In progress when handoff triggered
<exactly what was being worked on — file, function, test being fixed>
### Not started (known remaining work)
- <task not yet touched>
---
## Files in play
| File | Why it matters |
|------|----------------|
| `path/to/file` | <one line> |
---
## Key decisions made
### Decision: <title>
**Chose:** <what>
**Why:** <full reasoning — alternatives considered, why this was picked>
**Implications:** <what this constrains going forward>
---
## Failed approaches — do not retry
### Failed: <description>
**Tried:** <exactly what was attempted>
**Error / result:** <exact error or observed behavior>
**Why it failed:** <root cause>
**Alternative taken:** <what was done instead>
---
## Blockers and open questions
- [ ] <specific blocker — what is unknown and what would unblock it>
---
## Next steps
1. **<Action title>**
- File: `<path>`
- What to do: <specific instruction — enough to act without guessing>
- How to verify: <command or check that confirms success>
2. **<Action title>**
- ...
---
## Critical context
- **<title>:** <non-obvious thing that would cause mistakes if unknown>
---
## Environment at handoff
| Item | Value |
|------|-------|
| Language / runtime | `<e.g. Node 22.3.0>` |
| Package manager | `<npm / pip / cargo>` |
| Running services | `<port:service or "none">` |
| Relevant env vars | `<names only>` |
| Test command | `<exact command>` |
| Dev server command | `<exact command or N/A>` |
---
## Handoff metadata
| Field | Value |
|-------|-------|
| Handoff reason | manual request |
| Previous LLM | `<your model name and version>` |
| Tool | Claude Code |
| Generated | `<ISO 8601>` |
| Suggested next LLM | any |
---
*Generated by [llm-handoff](https://github.com/thatsuperdev/llm-handoff)*
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 · 222 lines · 16 tokens per session scan A cce05a402350
create is a command published in the GitHub repository thatsuperdev/llm-handoff (1 stars, last pushed 2mo ago), licensed MIT. It adds 16 tokens to every session and 1,415 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
assets
Inventory .claude-design/ / — file counts, types, locations.
start
Run the canvas-to-code workflow — guided discovery, auto-resume, or jump to a specific gate.
dashboard
Work timeline — what's done, what's in progress, when, with PR + doc links.
update
Pull the latest Canvas-to-Code into this repo — refreshes the plugin cache, even when the version field didn't change.
remembrall-help
List all Remembrall commands, skills, and config options.
phoenix
Toggle Phoenix mode on/off — recurring context recycling for zero-click continuity.