whats-next
Analyze the current conversation and create a handoff document for continuing this work in a fresh context
Skill by hhellbusch
Install
Works with Claude Code. Read the repository before you give any tool access to your data.
npx agentmarket install skill/hhellbusch-whats-next
Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.
Get the weekly diff
The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.
About
Use cases: documents
Reviews
Write a reviewNo reviews yet. Installed it? Reviews come from verified installers only. How reviews work
Permissions
What this skill can do and what it asks for, from its SKILL.md. Checked Oct 25, 2026.
- Tools
- No tools found
- Secrets
- None asked for
- Install scripts
- None found
Network hosts it talks to 1
codeload.github.com· found in install command
Files it touches
~/.claude/skills/whats-next
Tools it pre-approves 5
From allowed-tools in SKILL.md: the agent may use these without asking you each time.
ReadWriteShellWebSearchWebFetch
What's inside
- Skill name
whats-next- Description
- Analyze the current conversation and create a handoff document for continuing this work in a fresh context
- License
- MIT
- Allowed tools
- Read, Write, Shell, WebSearch, WebFetch
- Instructions
- 179 lines in SKILL.md · view on GitHub
Files (1)
- SKILL.md
SKILL.md (first lines)
# Whats-Next — Session Handoff Create a **compact** handoff so a fresh session can continue safely. Same file and project rules as `/checkpoint`; close-out may add framing assumptions and open threads. Not a full session transcript. Canonical fields: `kit/WORKING-STYLE.md` → Progressive bookkeeping → Continuation contract. ## Instructions **PRIORITY: Continuation over completeness.** Preserve the contract fields. Link durable detail. Expand only when the user asks or when safety / continuation requires it. ### Step 0: Evaluate whether a handoff is needed Before creating anything, check whether the session's work is already fully persisted: 1. Check `git status` and `git log --oneline -5` — is the working tree clean? Are all changes committed? 2. Is there any in-flight state — a half-finished task, a pending decision, context that only exists in the conversation? **If all work is committed and there's no in-flight state:** tell the user the session is fully persisted and no handoff is needed. If a stale `whats-next.md (project-scoped)` exists, ask whether to delete it. **If there is genuine in-flight state:** proceed to Step 1. --- ### Step 1: Capture continuation state Record only what the next session needs: 1. **Outcome** — what is true now (deliverable status; committed vs draft) 2. **Next action** — the immediate next step, with precise targets (paths, URLs) when known 3. **Decisions that bind** — trade-offs that would be re-litigated without knowing they were settled 4. **Boundaries** — constraints, authorization, safety limits 5. **Evidence pointers** — artifact paths + revisions, commits, dated observations; link rather than paste 6. **Checks and limits** — what was verified and what those checks do *not* prove 7. **Unresolved gate** — pending decision or blocker, if any 8. **Retry-relevant failures** — approaches that failed and would otherwise be retried; omit dead ends that are already obvious from git/docs 9. **Owner** — who holds the thread (default: this session’s human, unless delegated) Write to `.planning/<project>/whats-next.md` using the same project rule as `/checkpoint` and `/shoshin`: an explicit project name wins; one BRIEF is enough; several briefs and no name means ask and do not write. Do not pick the newest `BRIEF.md`. Append with a datestamp when the file already has content — do not replace. Before the narrative, record comparison fields a later `/shoshin` can check: current branch (`git branch --show-current`), short hash (`git rev-parse --short HEAD`), and UTC ISO time (`YYYY-MM-DDTHH:MM:SSZ`). --- ### Step 1.3: Spar trigger evaluation Before writing the handoff, evaluate the session output against these triggers. If **two or more** fire, surface a recommendation before proceeding: | Trigger | Signal | |---|---| | **Argumentative output** | An essay, proposal, brief, or plan was drafted or significantly revised this session | | **Decision with trade-offs** | A non-trivial technical or design choice was made (architecture, tool, scope, framing) | | **External-facing output** | Work is being prepared to share, publish, or submit | | **Unchallenged direction** | The session has produced uniformly agreeable outputs — no pushback, no alternatives considered | | **Load-bearing claim** | The work rests on a claim that wasn't verified or challenged during the session | | **Scope expansion** | The task broadened mid-session beyond what was originally requested | | **Surprising conclusion** | The session arrived at a conclusion that would have been non-obvious at the start | **If two or more triggers fire:** > "Spar conditions detected: [list which triggers]. Want to run `/spar` on [specific target] before closing? It would take ~5 minutes and this is the highest-value moment — the output is fresh and a spar here prevents carrying a weak argument into the next session." Do not run the spar automatically. Always ask first. If the session was purely mechanical — no arguments, no decisions, no output that could be wrong — skip entirely. --- ### Step 1.5: Shoshin — Assumptions check Before writing the handoff, identify assumptions this session is carrying that a fresh session should question rather than inherit: - What framing decisions were made in this session? - What was taken as given that might not be true? - Did the project's scope or framing shift during this session? If any assumptions are worth surfacing, include an `<assumptions_carried>` section in the handoff. If the session was straightforward with no framing decisions, skip it. --- ## Output Format Write `whats-next.md` using this structure. `<resume_compare>` is required. Keep sections short; omit empty optional blocks. ```xml <resume_compare> branch: [git branch --show-current] commit: [short hash] recorded: [YYYY-MM-DDTHH:MM:SSZ] external: [cluster/env/deploy note, or none] </resume_compare> <original_task> [Initial request — precise scope, one short paragraph] </original_task> <current_state> [Outcome now: deliverable status; committed vs draft; owner] </current_state> <next_action> [Immediate next step with precise targets; unresolved gate if any] </next_action> <decisions_and_boundaries> [Binding decisions; authorization/safety constraints; checks run and their limits] </decisions_and_boundaries> <evidence> [Links/paths + revisions or dated observations — not pasted dumps] </evidence> <retry_avoid> [Optional — failed approaches only if they would otherwise be retried] </retry_avoid> <assumptions_carried> [Optional — framing decisions a fresh session should question] </assumptions_carried> <open_threads>
From hhellbusch/zanshin-pi-extension, MIT.
More from hhellbusch/zanshin-pi-extension
- 0.565score
craft
hhellbusch
Apply engineering principles to code or design — DRY, KISS, SRP, YAGNI, convention over configuration, orchestration vs program, shared change scope, phased…
SkillAFreshness: AClaude Code +9 - 0.565score
domain-language
hhellbusch
Audit established domain terms across glossaries, ADRs, code, schemas, APIs, tests, UI, and external contracts.
SkillAFreshness: AClaude Code +9 - 0.565score
kaeshi
hhellbusch
Inversion lens (返し — reversal): what would guarantee failure?
SkillAFreshness: AClaude Code +9 - 0.565score
kihon
hhellbusch
Basics / fixed forms (基本): shell, secrets, vault, sql, git, k8s, testing, lint, ansible, helm, kustomize, python, typescript, structured edit.
SkillAFreshness: AClaude Code +9 - 0.565score
shoshin
hhellbusch
Surfaces load-bearing assumptions against source artifacts before building.
SkillAFreshness: AClaude Code +9 - 0.565score
unslop
hhellbusch
Cut AI tells from writing.
SkillAFreshness: AClaude Code +9 - 0.565score
yomi
hhellbusch
Second-order lens (読み — reading ahead): and then what?
SkillAFreshness: AClaude Code +9 - 0.559score
lab-journal
hhellbusch
Append one lab-journal entry after a meaningful surprise, or skip.
SkillAFreshness: AClaude Code +9 - 0.558score
checkpoint
hhellbusch
Save current session state mid-session — faster than /whats-next, designed for crash recovery
SkillAFreshness: AClaude Code +9 - 0.558score
spar
hhellbusch
Adversarial review of any content — essays, designs, plans, or ideas
SkillAFreshness: AClaude Code
Signals
- Stars
- 0
- Forks
- 0
- Open issues
- 0
- Last commit (this folder)
- Oct 5, 2026 (24 days ago)
- Last release
- Oct 5, 2026 (24 days ago)
From GitHub (fetched Oct 24, 2026).
Details
- License
- MIT
- First seen
- Oct 24, 2026
- Last checked
- Oct 24, 2026
- Source
- GitHubgithub-graphql
More from hhellbusch
- 0.565score
craft
hhellbusch
Apply engineering principles to code or design — DRY, KISS, SRP, YAGNI, convention over configuration, orchestration vs program, shared change scope, phased…
SkillAFreshness: AClaude Code +9 - 0.565score
domain-language
hhellbusch
Audit established domain terms across glossaries, ADRs, code, schemas, APIs, tests, UI, and external contracts.
SkillAFreshness: AClaude Code +9 - 0.565score
kaeshi
hhellbusch
Inversion lens (返し — reversal): what would guarantee failure?
SkillAFreshness: AClaude Code +9 - 0.565score
kihon
hhellbusch
Basics / fixed forms (基本): shell, secrets, vault, sql, git, k8s, testing, lint, ansible, helm, kustomize, python, typescript, structured edit.
SkillAFreshness: AClaude Code +9 - 0.565score
shoshin
hhellbusch
Surfaces load-bearing assumptions against source artifacts before building.
SkillAFreshness: AClaude Code +9 - 0.565score
unslop
hhellbusch
Cut AI tells from writing.
SkillAFreshness: AClaude Code +9
Cite this page
whats-next skill. Agent Mart. https://agentmart.dev/skills/hhellbusch-whats-next (last updated 2026-10-24)