agentsclimarketplace
Sign in

Handoff

Skill psenger/ai-agent-skills/skills/handoff

Make sure to use this skill whenever the user types /handoff (with or without a filename argument), says "handoff", "save state", "context rot", "save session", "create a handoff", "I need a clean start", wants to snapshot progress before clearing context or switching sessions, is approaching the context limit (300–400k tokens), or wants to delegate the current session state to another agent. Also invoke for RESUME mode: when the user says "load handoff", "resume from [file]", "continue where we left off", "pick up where I left off", "load the handoff at [path]", references a .claude/handoffs/ file path, or says there is a handoff file from a prior session. Creates or loads a structured JSON snapshot capturing goals, decisions, completed steps, pending work, constraints, and modified files so work can continue cleanly in a new session.From its SKILL.md

Install
npx -y skills add psenger/ai-agent-skills --skill handoff

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 9 stars9 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

3.5 KB, 678 tokens by cl100k_base, as published. Nobody here has run it

Handoff

Saves or loads a structured JSON snapshot of session state. Use the schema in references/schema.md for all fields.

Mode Selection

  • User wants to save / pause / context is full → CREATE workflow
  • User wants to resume / load / continue → RESUME workflow
  • After 5+ file edits or a major decision → suggest: "We've made good progress — say 'create handoff' when you're ready to save state and start fresh."

CREATE Workflow

Step 1 — Determine Output File

Use the user's argument if provided (e.g. /handoff auth-work.json). Otherwise default to .claude/handoffs/YYYY-MM-DD-HHMMSS-<slug>.json in the project root, where <slug> is 2–4 words from the task goal (e.g. auth-refactor, payment-api). Create .claude/handoffs/ if it doesn't exist.

Step 2 — Analyse the Session

Review the full conversation and all modified files, then extract:

  • The primary goal and what "done" looks like (acceptance criteria)
  • All architectural and design decisions reached, with rationale
  • What has been completed and can be verified
  • What still needs to be done, in priority order
  • Hard constraints and non-negotiables
  • Known bugs, issues, or unresolved warnings
  • Which files were changed and what was done to each

Step 3 — Write the Handoff File

Write the JSON using the schema in references/schema.md. Do not omit any top-level field — use [] if there is nothing to report.

Step 4 — Confirm and Guide the User

Handoff written to: <filename>

Summary:
- Task: <goal>
- Completed: <N> steps
- Pending: <N> steps
- Constraints: <N>
- Open issues: <N>
- Modified files: <N>

Next steps:
1. Run /clear to wipe the context window
2. Start a new session
3. Paste this prompt to resume:

"<resume_prompt value>"

Do not run /clear yourself — the user must do this manually.

RESUME Workflow

Step 1 — Locate the File

Use the user's argument if given. Otherwise read .claude/handoffs/ and pick the most recent file by filename timestamp. If the directory is empty or missing, tell the user.

Step 2 — Load and Orient

Read the file fully. Then report:

Loaded handoff: <filename>

- Task: <goal>
- Completed: <N> steps
- Pending: <N> steps
- First step: <pending_steps[0]>
- Open issues: <N>
- Constraints: <N>

Step 3 — Begin Work

Start immediately with pending_steps[0]. Keep constraints, discovered_issues, and decisions visible as you work — don't re-litigate choices already made.

Reference Files

  • references/schema.md — full JSON schema, field rules, and a complete example

What ships with it: 13 files

46.0 KB alongside SKILL.md

.workspace/

references/

Gives 0 of the 12 instructions most agent orchestration skills give in 678 tokens

Counted across 848 of the 1,300 authors here whose files we hold, read 2026-09-06

  • Dispatch one agent per independent problem domainin 56 of 848, across 42 files
  • Run full test suite after integrationin 55 of 848, across 42 files
  • Verify fixes do not conflictin 40 of 848, across 32 files
  • Review each summary when agents returnin 40 of 848, across 31 files
  • Write a handoff document summarising the current conversationin 30 of 848, across 25 files
  • Reference existing artifacts by path or URLin 26 of 848, across 24 files
  • Give each agent a specific scopein 19 of 848, across 10 files
  • Give each agent a clear goalin 19 of 848, across 10 files
  • Include a suggested skills section in the documentin 18 of 848, across 16 files
  • Tailor the doc to the user argumentsin 18 of 848, across 15 files
  • Issue all subagent dispatches in the same responsein 17 of 848, across 11 files
  • Use git worktrees for isolationin 17 of 848, across 8 files

Said here and by no other author read

  • Determine output file for handoff
  • Analyse the session
  • Confirm and guide the user
  • Load and orient
  • Begin work

Grouped from the skills themselves: near-identical wordings counted once, and counted by distinct author, so one author publishing three of these counts once. Length counted with cl100k_base; the agent that loads this file may tokenize it differently.

Keep looking

Skills are one crate of 325,949. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.