create

A handoff document for another AI coding agent. It records the task, current project state, completed work, remaining work, decisions, and approaches that failed.

In plain words
What is it for?
Use it to document repository status, recent changes, the goal, unfinished items, errors, and user preferences. The resulting HANDOFF.md file guides another agent continuing the task.
Why use it?
It prevents the next agent from having to reconstruct the work from scratch and preserves important context between sessions.

Command

Install

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.

agentmods
npx agentmods add commands/willseltzer/claude-handoff/create
Clone the repo
git clone --depth 1 https://github.com/willseltzer/claude-handoff
Per session 14 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 1,015 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What 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.

ModelPer sessionOnce invoked
Fable 5 $0.00014 $0.01015
Opus 5 $0.00007 $0.00508
Sonnet 5 $0.00003 $0.00203
Haiku 4.5 $0.00001 $0.00102

Measured 2d ago against content hash 6ffd3a02d373, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

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 2d ago.

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.

commands/create.md · 142 lines

How it starts

The opening of the file, as written. The whole thing — 142 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Create a HANDOFF.md file that enables ANY AI coding agent to continue this work seamlessly.

Gather Context First

Run these commands to understand current state:

  • git status - see uncommitted changes
  • git diff --stat - see what files changed
  • git log --oneline -5 - recent commits from this session

Review the conversation history to extract:

  • The original task/goal
  • What was completed
  • What was tried and didn't work (critical - saves hours)
  • Key decisions and their rationale
  • User preferences expressed during the session
  • Error messages encountered and how they were resolved

Write HANDOFF.md

Use this structure (omit empty sections, but NEVER omit Failed Approaches if any exist):

# Handoff: [Brief Task Title]

**Generated**: [date/time]
**Branch**: [git branch]
**Status**: [In Progress / Blocked / Ready for Review]

## Goal

[1-2 sentences: what the user wants to achieve]

## Completed

- [x] [Specific completed item]
- [x] [Another completed item]

## Not Yet Done

- [ ] [Remaining task - be specific]
- [ ] [Another remaining task]

## Failed Approaches (Don't Repeat These)

[IMPORTANT: Always include this if anything was tried and abandoned. Be specific:]
- What was attempted
- Why it failed (error message, performance issue, design flaw)
- Why the current approach is better

Example:
> Tried using passport.js for OAuth but it conflicted with existing Express middleware (req.user was undefined). Switched to oauth4webapi which works directly with fetch.

## Key Decisions

| Decision | Rationale |
|----------|-----------|
| [Choice made] | [Why this approach] |

## Current State

**Working**: [What's functional right now]

**Broken**: [What's not working, error messages if relevant]

**Uncommitted Changes**: [Summary of unstaged/staged changes]

## Files to Know

| File | Why It Matters |
|------|----------------|
| `path/to/key/file.ts` | [Brief description] |

## Code Context

[Include actual code the next agent needs. Don't describe - show:]

**Key interfaces/signatures** (so the agent knows how to call/modify them):
```typescript
// Example: hook signature
function useAuth(): { user: User | null; login: (creds: Credentials) => Promise<void> }

API request/response shapes (if backend work):

// POST /api/resource - example response
{ "id": 123, "status": "created" }

Non-obvious logic (anything tricky that isn't self-documenting)

Resume Instructions

[Be extremely specific. Not "test the feature" but step-by-step with expected outcomes:]

  1. [Setup step if needed - migrations, env vars, etc.]
  2. [First action with exact command or file to edit]
  3. [Verification step with expected outcome]
    • Expected: [what should happen]
    • If it fails: [what to check]

Example:

  1. Run alembic upgrade head to apply migrations
  2. Start server: ./start.sh
  3. Test login flow: POST to /api/login with [email protected] / testpass
    • Expected: 200 response with { token: "..." }
    • If 401: Check user exists in DB

Setup Required

[Only if there are prerequisites the next agent needs:]

  • Environment variables: API_KEY, DATABASE_URL
  • Test accounts: [email protected] / password123
  • Required services: Redis must be running on :6379

Edge Cases & Error Handling

[Document known edge cases and how they're handled - or should be:]

  • What happens if [X fails]? → [current behavior or "not handled yet"]
  • What if user does [Y]? → [expected behavior]

Warnings

[Gotchas, things that look wrong but are intentional, or traps to avoid]

Read the full file on GitHub · 142 lines

Changes

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.

  1. 2d ago First seen · 142 lines · 14 tokens per session scan A 6ffd3a02d373

Subscribe to this mod's changes

create is a command published in the GitHub repository willseltzer/claude-handoff (141 stars, last pushed 8mo ago), licensed MIT. It adds 14 tokens to every session and 1,015 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 2026-08-30.