agentsclimarketplace
Sign in

Handoff

Skill ananthanandanan/skills/skills/engineering/handoff

Generate a Markdown handoff document summarizing the current conversation so another agent (Claude, Codex, opencode, pi, ampcode, Cursor, or a fresh session) can resume without re-deriving context. Produces docs/<slug>-handoff.md. Invoke explicitly when a session is near its context limit.From its SKILL.md

Install
npx -y skills add ananthanandanan/skills --skill handoff

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

3 things to look at

  • skips confirmationTells the agent to proceed without asking first, 1 time: "do not ask interview questions".
  • 4 stars4 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.
  • runs commandsInstructs the agent to run 1 command, including `mkdir -p docs`.

SKILL.md

3.2 KB, 691 tokens by cl100k_base, as published. Nobody here has run it

handoff

This skill runs when context is nearly exhausted. Budget is the constraint. Draft from what's already in conversation context — do not re-read files, do not grep, do not ask interview questions, do not run a self-check pass. One Write call is the goal.

Output path

docs/<slug>-handoff.md where <slug> is a 2–3 word kebab-case identifier derived from the session's main agenda (or the user's freeform argument to the slash command, if given).

  • Lowercase, alphanumeric + hyphens only.
  • If the file exists, append -2, -3, … — never clobber.
  • mkdir -p docs if needed.

Audience

Another LLM coding agent with Read/Write/Bash but zero prior context. Write for them: concrete paths, exact commands, exact error strings. No prose pleasantries, no recap of the dialogue.

Sections (in order)

Include only sections with real content. Omit or write None this session. for empty ones — never pad.

  1. Header — one-line crux (what this session was about), branch, date, working dir.
  2. Agenda — 2–4 sentences on the goal.
  3. Files read — bullet list: path + half-sentence on why it mattered.
  4. Files written / edited — table: File | Change.
  5. Bugs found — symptom + root cause (if known) + file:line.
  6. Fixes made — what was broken + the fix + where.
  7. Skills used — /ank:<skill> invocations + outcome. Omit if none.
  8. Links / references — URLs, issue/PR numbers, doc paths.
  9. Key decisions — non-obvious choices + why. Prevents re-litigation.
  10. Current state — what works, what's half-done, what's untested. Quote actual build/test output if you ran any; say so if you didn't.
  11. Open questions / not done — explicit gaps, deferred items.
  12. Next steps — concrete, ordered. Each item must be actionable without a clarifying question (e.g. "read frontend/lib/auth.ts and trace refreshToken()" — not "figure out auth"). Write this last.

Optional (include only if substantial and quick to produce): a fenced ASCII layout/architecture diagram, key types/state shape, or API surface touched.

Shape: tables for file lists, fenced ASCII for diagrams, short headers, no filler prose.

Constraints

  • Markdown only. No HTML, images, or external assets.
  • No invented paths, line numbers, commits, or counts. If unsure, omit — don't verify, you don't have the budget. The receiving agent will grep.
  • No transcript dump. Distill the work, not the dialogue.
  • No interview. Use the user's slash-command argument if given; otherwise derive the slug and crux yourself.
  • No self-check pass. Get the file written.
  • Don't run builds, tests, or migrations to populate the doc. Report state as observed.

What ships with it

Read from the repository

Just SKILL.md. No reference files, no scripts.

Gives 0 of the 12 instructions most docs writing skills give in 691 tokens

Counted across 1,951 of the 3,904 authors here whose files we hold, read 2026-09-06

  • Use third-person for skill descriptionsin 54 of 1951, across 35 files
  • Start descriptions with Use whenin 43 of 1951, across 29 files
  • Run baseline scenarios before writing any skillin 40 of 1951, across 26 files
  • Use active voicein 40 of 1951, across 36 files
  • Map file responsibilities before defining tasksin 36 of 1951, across 29 files
  • Use checkbox syntax for tracking stepsin 35 of 1951, across 27 files
  • Ask one question at a timein 35 of 1951
  • Offer execution options after saving the planin 33 of 1951, across 24 files
  • Include complete code in every stepin 33 of 1951, across 27 files
  • Design units with clear boundaries and interfacesin 31 of 1951, across 23 files
  • Announce the skill usage at the startin 30 of 1951
  • Verify agent compliance after adding the skillin 29 of 1951, across 17 files

Said here and by no other author read

  • Draft handoff from existing conversation context
  • Write for an LLM with no prior context
  • Include only sections with real content

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.