handover

A written handoff prompt that records the decisions, limits, confirmed findings, rejected options, and current progress from one coding-agent session. A new agent can use it to continue work without the earlier conversation.

In plain words
What is it for?
Use it when unfinished work must move to a fresh agent session. It documents what is complete, partly complete, or still pending, but does not implement the described work.
Why use it?
Important context can disappear when a session ends or its context window fills up. The handoff keeps the next session from repeating research or violating settled constraints.

Skill for Claude CodeCodex

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 skills/fmind/dotfiles/handover
Any agent
npx skills add fmind/dotfiles --skill handover
Clone the repo
git clone --depth 1 https://github.com/fmind/dotfiles

Made for: Claude Code, Codex.

Per session 38 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 1,372 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 $0.00038 $0.01372
Opus 5 $0.00019 $0.00686
Sonnet 5 $0.00008 $0.00274
Haiku 4.5 $0.00004 $0.00137

Measured yesterday against content hash df3cbb517998, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

handover 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.

skills/handover/SKILL.md · 73 lines

How it starts

The opening of the file, as written. The whole thing — 73 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Handover

Turn everything this session established into one prompt a fresh agent can execute without asking a question.

Authority Boundary

Writing a handover is a documentation act. Produce the prompt file and stop. Do not implement the work it describes, commit it, or push it unless the user separately asks. The next session owns execution.

What a Handover Carries

The receiving agent starts with an empty context window. It sees the repository and the prompt — nothing else. Four classes of knowledge die with this session unless the prompt captures them:

  • User-stated constraints: preferences, rejections, and scope limits the user gave in conversation. These are authoritative and unrecoverable from the code.
  • Verified findings: facts confirmed this session by reading dependency source, running a command, or fetching documentation — with the evidence that settled them.
  • Rejected alternatives: options considered and dropped, with the reason. Without these the next session re-litigates settled decisions or walks into a known dead end.
  • Current position: what is done, what is half-done, and what has not started.

Repository state that the next agent can read for itself is not knowledge — it is noise. Point at files instead of transcribing them.

Workflow

  1. Re-read the session: Recover the original request, every subsequent user correction, and the decisions each one settled. A later correction overrides an earlier instruction; carry the resolved position, not the debate.
  2. Separate instruction from inference: Mark what the user required, what you proposed and they approved, and what you assumed without confirmation. The next agent must know which lines it may revise.
  3. Verify before you write: Re-check every path, symbol, command, and version the prompt will name. Stale coordinates are worse than absent ones — they send the next session confidently to the wrong place. Drop or re-derive anything you cannot confirm now.
  4. Capture the position: Record completed work with its proof, in-progress work with its exact stopping point, and untouched work. Name the branch and whether the tree is clean.
  5. Record the trade-offs: For each real decision, state the choice, the alternatives, and why. For each dead end, state what failed and how it failed, so the next session does not retry it.
  6. State the proof contract: Name the commands that must pass before the work is done — normally mise run check and mise run test — and any check that is currently red.
  7. Draft as an instruction, not a report: Address the receiving agent in the imperative. It needs orders and context, not a narrative of what you did.
  8. Test for self-containment: Read the draft as someone who never saw this session. Every pronoun must resolve, every "as discussed" must be replaced by the substance, every file reference must be a path. Rewrite anything that only makes sense to you.
  9. Write the file: Save to .agents/prompts/<YYYY-MM-DD>-<slug>.md and report the path so the user can paste or pipe it into the next session.

Read the full file on GitHub · 73 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. yesterday First seen · 73 lines · 38 tokens per session scan A df3cbb517998

Subscribe to this mod's changes

handover is a skill published in the GitHub repository fmind/dotfiles (4 stars, last pushed 2d ago), licensed MIT. It adds 38 tokens to every session and 1,372 once invoked, about $0.0002 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.

Related

Other skills, from other repositories

dotfiles-bootstrap

Bootstrap a workstation with the dotfiles framework. Takes a GitHub user / owner+repo / explicit clone URL and runs dot init (which shells out to chezmoi) with the right safety prompts. Honors the active agent profile (ask / plan / apply / audit) so it defaults to dry-run in safer modes and full apply in apply.

sebastienrousseau/dotfiles · 88 tokens

vibe

Delegate a coding task to a cheap AI model (Mistral Vibe by default, but any provider Vibe knows about — DeepSeek, Gemini Flash, etc.) and supervise the result via git diff. Claude orchestrates, the cheap model codes. Claude consumes 500-1500 tokens per delegation regardless of how many file reads the delegate does…

sebastienrousseau/dotfiles · 137 tokens

aiq-research

Use when asked to run deep research or AI-Q research through a reachable NVIDIA AI-Q Blueprint backend.

laurigates/dotfiles · 25 tokens

obsidian-bases

Obsidian Bases database feature for YAML-based interactive note views. Use when creating .base files, writing filter queries, building formulas, configuring table/card views, or working with Obsidian properties and frontmatter databases.

laurigates/dotfiles · 49 tokens

telegram

Send notifications, interactive questions, or multiple-choice polls to the user via Telegram. Use when the user asks to be notified ("ping me", "notify me on Telegram", "ask me when..."), when a long-running task finishes and the user is likely away, when an irreversible action needs out-of-band confirmation, or when…

laurigates/dotfiles · 117 tokens

chezmoi-expert

Comprehensive chezmoi dotfiles management expertise including templates, cross-platform configuration, file naming conventions, and troubleshooting. Covers source directory management, reproducible environment setup, and chezmoi templating with Go templates. Use when user mentions chezmoi, dotfiles, cross-platform…

laurigates/dotfiles · 88 tokens