handoff
Skill0 starsChanged 4 months agoMITView on GitHub
Generate a structured continuity handoff on demand — before a /clear, at session end, when handing to a fresh agent/subagent, or when passing work to the other machine (aiserver↔gcloud). Picks the right destination + format automatically. Use when Nile says "hand off", "write a handoff", "/handoff", "prep for /clear", "pass this to the gcloud box / aiserver", or "summarize this session so the next one can continue". Inherits the repeatable-generator + suggested-skills + reference-don't-duplicate + redaction discipline from the mattpocock/skills handoff idea, re-derived in Nile's richer format. On-demand: suggest it, don't auto-fire.
- Reads credentials
191 other projects use a version of this.
Adopted by
Most recently taken up by
- tanishq286/tq-forgetheir own version, first seen Nov 15, 2026
- niki914/zafirotheir own version, first seen Nov 15, 2026
- davidondrej/skillstheir own version, first seen Nov 15, 2026
- quickchatai/quickreviewtheir own version, first seen Nov 15, 2026
- HiteSit/real-scientist-skillstheir own version, first seen Nov 15, 2026
- alexj11324/Mikkatheir own version, first seen Nov 15, 2026
Used with
Projects with it also use decision-mapping, setup-ts-deep-modules and 10 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.
FabienDehopreSkilldecision-mapping
Turn a loose idea into a sequenced map of investigation tickets, then drive them to resolution one at a time.
Also in 8 projects that use this
liufeng1976Skillsetup-ts-deep-modules
Wire dependency-cruiser into a TypeScript repo so each package is a deep module — implementation hidden in subfolders, reachable only through its entry-point files. User-invoked.
Also in 42 projects that use this
AWSSkillto-questionnaire
Turn a decision you can't fully answer into a questionnaire for someone else to fill in.
Also in 93 projects that use this
GoogleSkillask-matt
Ask which skill or flow fits your situation. A router over the skills in this repo.
Also in 87 projects that use this
RoentekSkillloop-me
Grill me about specs for the workflows I want to build, within this workspace.
Also in 51 projects that use this
LesaniSkillwriting-beats
Shape an article as a journey of beats, choose-your-own-adventure style. The user picks a starting beat from the raw material, you write only that beat, then offer options for where to pivot next, beat by beat, until the article reaches a natural end. Use when the user has raw material and wants to assemble it as a narrative rather than an argument.
Also in 54 projects that use this
Other versions191 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 191
- mattpocock/skills★ 325ksame name
- QuivrHQ/quivr★ 40ksame name
- The-Vibe-Company/quivr★ 40ksame name
- GreyDGL/PentestGPT★ 16ksame name
- ykdojo/claude-code-tips★ 10ksame name
- ArcReel/ArcReel★ 6.6ksame name
- vinvcn/mattpocock-skills-zh-CN★ 5.8ksame name
- withcoral/coral★ 4.9ksame name
- epicenter-md/epicenter★ 4.8ksame name
- EpicenterHQ/epicenter★ 4.8ksame name
181 more not listed
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository9 other files it gives its agents.
- adhd-friendlySkill
- atom-cypher-safeSkill
- document-mind-mapSkill
- interview-modeSkill
- notebooklm-bulk-labelSkill
- prompt-forgeSkill
- skill-scoutSkill
- what-should-i-automateSkill
- zero-as-answerSkill
The file
We split this skill into the rules it gives an agent
6 rules. 0 are common, 1 is distinctive, and 5 are very specific to this project.
0
Common
Most of the 172 edited versions we read keep these.
1
Distinctive
This author's own choices. Worth a look.
5
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 30
If Nile passed an argument, treat it as "what the next session will focus on" and tailor the whole doc to that (lead with what's relevant …
- Very specific: line 39
Redact secrets — hard rule: Never write token/secret *values*.
- Very specific: line 45
Update ~/Exocortex/hot_cache.md: replace/refresh the ### ⮕ CONTINUE-FROM block near the top (keep it TIGHT — the file is a ~500-word …
- Very specific: line 55
Don't bloat hot_cache — if there's heavy detail, it belongs in a Todoist task/Drive doc that you *reference* here.
- Very specific: line 59
Write ~/.claude/gcloud-handoff/<ISO8601-compact>Z-handoff.md (e.g.
- Very specific: line 86
Write to ${TMPDIR:-/tmp}/handoff-<ISO>Z.md, report the path.
“Common” is measured: kept by at least 6 in 10 of the 172 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
handoff — Structured Continuity Generator
Generate a handoff doc that lets the next context — a fresh session here, a subagent, or the other machine — pick up exactly where this one left off, with no re-derivation. Your handoffs were hand-authored ad-hoc; this makes generation deterministic, complete, and consistently formatted.
It does not replace the existing continuity layers — it's the missing generator that writes into them correctly:
hot_cache.md(~/Exocortex/) — same-machine, session→next-session rolling cache (auto-maintained; this skill updates its CONTINUE-FROM block, never a parallel file).gcloud-handoff/(~/.claude/) — cross-machine peer transfer, ingestion-ready memory-entry blocks, archived.- capture buffer — inline→canon merge (separate concern; reference it, don't duplicate it).
Step 0 — pick the mode (destination decision tree)
Infer from context; ask only if genuinely ambiguous.
Who picks this up next?
├─ The next session on THIS machine (before /clear, or session end)
│ → MODE A: update hot_cache CONTINUE-FROM
├─ The OTHER machine (aiserver ↔ gcloud satellite)
│ → MODE B: write ~/.claude/gcloud-handoff/<ISO>Z-handoff.md (memory-block format)
└─ A throwaway fresh agent / subagent / external paste (no persistence needed)
→ MODE C: write to a temp path ($TMPDIR or /tmp), report the pathIf Nile passed an argument, treat it as "what the next session will focus on" and tailor the whole doc to that (lead with what's relevant to that focus; drop unrelated detail).
Universal rules (every mode)
These are the quality layer — they apply regardless of destination:
- Lead with the
role:line. State the role the next context should assume (executor | architect | researcher | auditor, per SOUL.md). Defaultexecutorunless the work implies otherwise. - Reference, don't duplicate. If something is already captured — a Todoist task (give the ID), a Drive doc, a KB note, the capture buffer, a commit, a memory file — link/cite it by path or ID, do not re-paste its content. The handoff is a pointer-map + the in-flight state that isn't written down anywhere else yet. This is the anti-bloat rule; a handoff that re-paginates existing artifacts is a bad handoff.
- Suggested skills / next actions. Include a short section naming the skills the next context should invoke and the immediate next steps — the forward-pointer. (e.g. "→ resume with
/change-disciplineon the pending vault sweep; the dry-run is at …".) - Redact secrets — hard rule. Never write token/secret values. Reference GSM secrets by NAME only (
YNAB_ACCESS_TOKEN), never the value. No API keys, passwords, or PII (this is outward-portable; apply the secrets + sensitive-data discipline). Scrub before write. - State the in-flight state honestly. What's done, what's mid-flight, what's blocked and on what. If a background task is running, say so and where its output lands. If something's unverified, mark it.
- Timestamp + one-line. Open with a one-line "what this session did" so the reader orients in 5 seconds.
Mode A — next-session (same machine) → hot_cache CONTINUE-FROM
Update ~/Exocortex/hot_cache.md: replace/refresh the ### ⮕ CONTINUE-FROM block near the top (keep it TIGHT — the file is a ~500-word rolling cache; older arcs roll to hot_cache_archive.md via hotcache-archive-retain.py, don't fight that). Match the existing shape:
### ⮕ CONTINUE-FROM (<YYYY-MM-DD> — <one-line focus>) — READ FIRST
**role: <role>.**
- <in-flight state, one bullet each — reference task IDs / paths, don't paste>
- **Suggested next:** <skills to invoke + immediate steps>
- **Running bg (check if no report):** <any background agents + where output lands>Don't bloat hot_cache — if there's heavy detail, it belongs in a Todoist task/Drive doc that you reference here. Update the file's # Last updated: header line too.
Mode B — cross-machine → gcloud-handoff (ingestion-ready)
Write ~/.claude/gcloud-handoff/<ISO8601-compact>Z-handoff.md (e.g. 20260615T161748Z-handoff.md). Match the exact existing schema — the receiving machine's boot-sweep (cortex-inbox-check.sh) flags it and the next session ingests→integrates→archives it, so the format must parse:
# Handoff — <short title of what this session accomplished>
**One-line:** <2-3 sentence summary>
**role:** <role>
---
## Memory entry 1 (<project|reference|feedback|user>)
- **name:** <kebab-slug>
- **description:** <one-line, used for recall relevance>
- **type:** <project|reference|feedback|user>
<self-contained body — for cross-machine, entries ARE self-contained (the peer may not have your local files), but still reference shared canon (Drive/Todoist IDs) the peer can also reach.>
---
## Memory entry 2 (...)
...Machine-agnostic: works aiserver→gcloud OR gcloud→aiserver — don't assume which is hub (the aiserver-decommission is in progress; the direction is shifting). Name the source machine in the title. Archive prior handoffs per the existing archive/ convention only after they're confirmed ingested.
Mode C — throwaway / fresh agent
Write to ${TMPDIR:-/tmp}/handoff-<ISO>Z.md, report the path. Use the article's lean shape (one-line + role + state + suggested-skills + references). No persistence, no archive — it's for an agent that needs context once. Still redact.
Anti-patterns (don't)
- Don't re-paste what's already written. Reference task IDs / paths / commits. The buffer, hot_cache, Todoist, Drive, and memory files are the canon — point at them.
- Don't write a parallel same-machine artifact. Mode A updates hot_cache; it doesn't create
handoff-local.md. One continuity surface per machine. - Don't break the gcloud-handoff schema. The peer's ingest expects the memory-entry-block format. Malformed → the next session can't integrate it.
- Don't leak secrets. GSM names only, never values. Scrub before write — this artifact is portable and may cross machines.
- Don't bloat. A handoff is a pointer-map + in-flight state, not a transcript. If it's getting long, you're duplicating artifacts that should be referenced.
Provenance
Native re-derivation (inherit→atomize→rebuild) of the mattpocock/skills handoff idea (surfaced via the Firecrawl best-skills list, 2026-06-15), vetted against Nile's existing 3-layer continuity infra and 3×3×-tested. Inherited from the article: repeatable-generator, suggested-skills forward-pointer, reference-don't-duplicate, redaction, argument-tailoring. Kept from Nile's setup: the role: line, memory-entry-block format, hot_cache CONTINUE-FROM convention, gcloud-handoff persistence/archive. 3×3× refinements folded in: same-machine updates hot_cache (no parallel artifact), match the gcloud-handoff schema exactly (ingest safety), machine-agnostic (aiserver-decommission in progress). Composes with cortex-inbox-check.sh (boot-sweep) + hotcache-archive-retain.py (rolloff).
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
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.

