agentleFS
GerardoRdz96GerardoRdz96/claude-skills/skills/session-handoff/SKILL.md

session-handoff

Skill0 starsChanged 4 months agoMITView on GitHub

Use when Gera wants to wrap up, clear context, switch machines or tools, or continue work in a fresh session — "hand off", "wrap up this session", "save where we are", "I need to clear context", "write a handoff", "we're at the token ceiling", "/session-handoff". Prints AND saves a structured handoff note (started / decisions locked / shipped / key files / running state / verification / deferred / next steps / gotchas) so the next session — in Claude Code, a Cowork tab, Codex, Gemini, or any tool — picks up exactly where this one left off without re-deriving anything. The cheapest insurance an AIOS has against context loss.

139 other projects use a version of this, 1 of them edited.

Adopted by

Most recently taken up by

Used with

Projects with it also use ascii-diagram, research-fleet and 8 more.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • SethGammonSkill

    ascii-diagram

    Generate perfectly aligned ASCII diagrams — architecture, flow, sequence, box-and-arrow. Uses a programmatic character-grid approach so alignment is guaranteed by math, not token prediction. Includes post-render verification.

    Also in 4 projects that use this

  • SethGammonSkill

    research-fleet

    Parallel research using Fleet wave mechanics. Spawns multiple scout agents, each investigating a different angle of the same question. Findings are compressed between waves. Produces a unified research brief from multiple independent perspectives.

    Also in 3 projects that use this

  • vicoa-aiSkill

    live-preview

    Start a local app in the current user project, expose it through a tunnel, and return preview details in structured Markdown. Use when the user asks for a preview URL, public URL, shareable localhost, mobile/web preview, or tunnel setup.

    Also in 4 projects that use this

  • botlearn-aiSkill

    doc-gen

    Also in 4 projects that use this

  • flyingobsidianSkill

    terminology

    Read a codebase, find terminology, create/update definitions into a markdown document.

    Also in 3 projects that use this

  • Dizzy-TenSkill

    bug-diagnosis

    Diagnose bugs and product issues without modifying code. Use when the user explicitly invokes $bug-diagnosis or asks for diagnosis, root-cause analysis, issue classification, or fix strategy for code bugs, failing tests, configuration/dependency problems, deployment issues, performance regressions, data anomalies, or frontend/user-experience problems described through screenshots, annotations, logs, errors, or natural-language reports. Prioritize analysis and recommendations only; do not edit files.

    Also in 3 projects that use this

  • +1 more from SethGammon

The file

We split this skill into the rules it gives an agent

3 rules. 0 are common, 1 is distinctive, and 2 are very specific to this project.

  • 0

    Common

    Most of the 117 edited versions we read keep these.

  • 1

    Distinctive

    This author's own choices. Worth a look.

  • 2

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    If a decision was made this session and isn't logged → offer to add it to decisions/log.md.

    line 70
  2. Very specific:

    Don't write to references/log.md unless the session actually ingested/shipped wiki work (that's the wiki timeline, not a session journal).

    line 71
  3. Distinctive:

    Report honestly: If something is half-done, broken, or unverified, the handoff says so plainly.

    line 75

“Common” is measured: kept by at least 6 in 10 of the 117 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

What this skill does

Generates a tool-agnostic handoff note for the current session: one markdown block a fresh session can read to continue the work cold. Servy is multi-LLM (Claude Code + Cowork + Codex + Gemini + local brains), so the note must orient any of them — name files by full path, never "the file we edited".

Doctrine: references/claude-code-context-management.md (#22 — the ~120k-token ceiling: cross ~12% of the 1M window → ask for a full summary → /clear → paste it back; #23 session chaining: discovery → planning → execution as separate warm-started sessions). Pattern credit: Nate Herk — "if you keep retyping the same prompt, that's a skill."

When to fire

  • Gera says he's done for now, switching tools (Claude Code ↔ Cowork ↔ Codex/Gemini ↔ another machine), or context is filling.
  • Proactively offer one line (don't insist) when a long session is clearly winding down, or when /context / the statusline shows the thread past the fire band: ~120k tokens (conservative Servy default) — 250-300K on 1M-window models (Nate's ~25%-of-window rule, 2026-07-11 Nate 6h course [2:00:48]).
  • New-build vs same-build key (Nate [5:19:16]): starting a different project/build while context is high → hand off for a fresh window; still editing and improving the same build → stay in the session, the warm context is the asset.
  • A 3rd–4th /compact is about to be needed — quality degrades past 3–4 compacts in a row [5:38:39], hand off instead.
  • Gera is stepping away for more than ~1 hour (the subscription cache TTL): the next message will reprocess everything cold anyway, so bank the state now [5:52:54].

Sibling boundary — pick the right tool

  • /session-handoff (this skill) — general, tool-agnostic, any session. Use for "wrap up / switch tools / clear context / save where we are". Output is a portable note + a saved file.
  • /gsd-pause-work — mid-GSD-phase pause that writes GSD's own state files; use only inside an active GSD phase. /gsd-resume-work restores it.
  • They compose: inside a GSD phase that's also ending the day, run /gsd-pause-work for the phase state, then /session-handoff for the human-readable cross-tool note.

The handoff template

Produce this filled in from the ACTUAL session — concrete paths and command-ready next steps, never vague summaries. Drop any section that's genuinely empty rather than padding it.

# Session handoff — {YYYY-MM-DD} — {topic}

## Where we started
- {the goal/request that opened the session}

## Decisions locked (don't relitigate)
- {decision} — why: {one line}  ({link to decisions/log.md if logged})

## Shipped this session
- {completed item, with file paths + commit SHA if committed/pushed}

## Key files (the map for a cold start)
- {path} — {what it is / why it matters now}

## Running state (live as of handoff)
- Branch: {branch} · pushed: {yes/no} · build/tests: {green/red/last result}
- Background work: {any routines armed, agents/workflows mid-run, deploys in flight}

## Verification deferred
- {anything shipped but NOT verified end-to-end — what's untested and the command/check that would verify it}

## In progress / next steps
1. {next concrete action, specific enough to start cold — include the command if there is one}

## Waiting on Gera (gates)
- {decision or manual step only he can do}

## Gotchas for the next session
- {anything surprising: a failing test, a quirk found, a constraint, an env caveat}

Where it goes

  1. Print it in chat (always — Gera may just copy it into the next session). The resume path is /copy → /clear → paste into the fresh session: takes under a minute, vs a slow /compact — and autocompact fires too late to save quality anyway (Nate 6h course, 2026-07-11 [5:53:25]).
  2. Save it to knowledge/handoffs/{YYYY-MM-DD}-{slug}.md (create the folder if needed). knowledge/ is the immutable drop zone — a handoff is a session artifact, it lives here, not in the wiki.
  3. Reconcile the trackers (don't duplicate — point):
    • Any follow-up that would otherwise be lost → append to pending.md and say you did.
    • If a decision was made this session and isn't logged → offer to add it to decisions/log.md.
    • Don't write to references/log.md unless the session actually ingested/shipped wiki work (that's the wiki timeline, not a session journal).

Guardrails

  • Report honestly. If something is half-done, broken, or unverified, the handoff says so plainly. A flattering handoff is a useless handoff — the next session inherits the lie.
  • Tool-agnostic. No "the function we just wrote" / "that file" — absolute paths and real names, because the reader may be Codex or a cold Claude with zero history.
  • Under a page. Orientation, not a transcript. If it's growing past a screen, you're journaling, not handing off.
  • Don't act on the work — this skill writes the note and reconciles trackers; it does not keep building. Wrapping up means wrapping up.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.