handoff
Use to compact the current session into a standalone handoff document so a FRESH agent (or you, later) can resume the work without the transcript.
Skill by KSriki
Install
Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.
npx agentmarket install skill/ksriki-handoff
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, design, writing, agent-building
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/handoff
What's inside
- Skill name
handoff- Description
- Use to compact the current session into a standalone handoff document so a FRESH agent (or you, later) can resume the work without the transcript. Triggers on "hand this off", "running out of context", "write a handoff", "continue this in a new session", "compact this for another agent", "pass this to the next agent". Optionally takes what the next session should focus on. Do NOT use to dump the full transcript, to write a design record (`write-design-doc`), a PR description (`review-pr`), or a PRD (`write-a-prd`) — a handoff REFERENCES those, it does not replace them.
- License
- MIT
- Instructions
- 73 lines in SKILL.md · view on GitHub
Files (1)
- SKILL.md
SKILL.md (first lines)
# Handoff: compact the session for a fresh agent > Adapted from `mattpocock/skills` (MIT) — reimplemented in this repo's style. ## When to use this You're ending a session, running low on context, or passing work to another agent, and the next one starts with **none of this conversation**. A good handoff is what lets them resume in one read instead of re-deriving everything. **Not this skill if:** you're recording a decision → `write-design-doc`. Opening a PR → `review-pr`. Writing a PRD → `write-a-prd`. A handoff points *at* those artifacts; it doesn't duplicate them. ## What a handoff is (and isn't) > **Reference, don't restate.** The handoff links to the spec, the ADR, the issue, > the commits, the diff — by path or URL. It is **not** a transcript dump and not a > second copy of artifacts that already exist. ## Procedure ### 1. Summarize for continuation, not for the record Write it forward-looking: **the goal, current state, decisions already made (with a one-line why), what's in flight, and the immediate next step.** Enough for a cold agent to act, no more. ### 2. Reference existing artifacts by path/URL Specs, design docs, issues, commits, diffs — link them, don't paste them. Restating an artifact means two copies that drift. ### 3. Add a "suggested next skills" section Name the skills the next agent should reach for (e.g. "work the next ticket with `tdd`; verify via `run-tests`; ship via `ci-cd`"). This is what makes the handoff *actionable* rather than just informative. ### 4. Redact sensitive data **Strip API keys, tokens, passwords, and PII** before writing. A handoff is a document that gets passed around — treat it as such (this is guardrail #2 in practice: secrets don't leak into artifacts). ### 5. Tailor to the next focus, if given If a next-session focus was specified, orient the doc around it — foreground what that focus needs, background the rest. ### 6. Write it to scratch, not into the repo Save to a scratch/temp location 〈your OS temp dir · or `docs/handoffs/` if you want it tracked〉. Don't commit it into the codebase unless that's the intent. ## Failure modes | Symptom | Cause | Fix | |---|---|---| | Next agent still lost | Summarized the past, not the next step | Write it forward-looking (goal → state → next) | | Two drifting copies of a spec | Restated an artifact instead of linking | Reference by path/URL only | | Secret ended up in the handoff | No redaction pass | Step 4 — strip keys/tokens/PII | | Handoff is a transcript wall | Dumped the conversation | Compact to decisions + state + next; link the rest | ## Definition of done - [ ] A **standalone, forward-looking** handoff doc a cold agent can act on - [ ] Existing artifacts **referenced by path/URL**, not duplicated - [ ] A **suggested-next-skills** section - [ ] **Redacted** — no keys, tokens, or PII - [ ] Tailored to the stated next focus (if any); saved to scratch, not committed by default
From KSriki/agent-harness-template, MIT.
More from KSriki/agent-harness-template
- 0.565score
evolve-harness
KSriki
Use to turn a repeated correction, a recurring failure, or a researched new practice/trend into a durable change to the AGENT LAYER — a new or edited skill…
SkillAFreshness: AClaude Code +9 - 0.524score
grill-me
KSriki
Use to grill the user relentlessly about a plan, decision, or idea — a frontier-batched interview down the design tree (rounds of ready questions, each with a…
SkillBFreshness: BClaude Code +9 - 0.524score
eval-harness
KSriki
Use when building, running, or debugging evaluations for LLM or agent components: golden sets, regression evals, LLM-as-judge, rubric scoring…
SkillBFreshness: BClaude Code +9 - 0.524score
observability
KSriki
Use when instrumenting a service (structured logs, metrics, traces), defining SLOs or alerts, building a dashboard, or triaging a live production incident /…
SkillBFreshness: BClaude Code +9 - 0.524score
orchestrate-agents
KSriki
Use to ship a multi-part change by running several agents IN PARALLEL — fan-out across subagents, each isolated in its own git worktree, then verify-and-merge.
SkillBFreshness: BClaude Code +9 - 0.524score
review-pr
KSriki
Use when preparing your own change for review (self-review, commit hygiene, writing the PR description) or when reviewing someone else's pull request for…
SkillBFreshness: BClaude Code +9 - 0.524score
ci-cd
KSriki
Use when setting up or debugging a CI pipeline, adding a workflow or gate, promoting a build through environments, deploying, cutting a versioned release, or…
SkillBFreshness: BClaude Code +9 - 0.524score
codebase-design
KSriki
Use to design or improve a single MODULE's interface — deciding where a seam goes, making code more testable / AI-navigable, or judging whether a module is…
SkillBFreshness: BClaude Code +9 - 0.524score
prd-to-issues
KSriki
Use to break a plan, spec/PRD, or the current conversation into a set of tracer-bullet tickets — each a narrow-but-complete vertical slice, declaring its…
SkillBFreshness: BClaude Code +9 - 0.520score
graphify
KSriki
Use for STRUCTURE and relationship questions about a codebase/corpus — how X reaches Y, what depends on what, blast radius, architectural hubs — and to build…
SkillBFreshness: BClaude Code +9 - 0.499score
harness-help
KSriki
Use when the user asks how to customize the harness or where something goes: "how do I add a rule / gotcha / skill", "where do I change the gate", "what file…
SkillBFreshness: BClaude Code +9 - 0.474score
debug-research
KSriki
Use when the answer likely lives OUTSIDE this repo: debugging an error from a library/framework/API, checking whether an SDK signature is current, evaluating…
SkillCFreshness: CClaude Code +9
Signals
- Stars
- 0
- Forks
- 0
- Open issues
- 0
- Last commit (this folder)
- Jul 22, 2026 (3 months ago)
From GitHub (fetched Oct 24, 2026).
Details
- License
- MIT
- Repository
- github.com/KSriki/agent-harness-template
- First seen
- Oct 24, 2026
- Last checked
- Oct 24, 2026
- Source
- GitHubgithub-graphql
More from KSriki
- 0.565score
evolve-harness
KSriki
Use to turn a repeated correction, a recurring failure, or a researched new practice/trend into a durable change to the AGENT LAYER — a new or edited skill…
SkillAFreshness: AClaude Code +9 - 0.524score
grill-me
KSriki
Use to grill the user relentlessly about a plan, decision, or idea — a frontier-batched interview down the design tree (rounds of ready questions, each with a…
SkillBFreshness: BClaude Code +9 - 0.524score
eval-harness
KSriki
Use when building, running, or debugging evaluations for LLM or agent components: golden sets, regression evals, LLM-as-judge, rubric scoring…
SkillBFreshness: BClaude Code +9 - 0.524score
observability
KSriki
Use when instrumenting a service (structured logs, metrics, traces), defining SLOs or alerts, building a dashboard, or triaging a live production incident /…
SkillBFreshness: BClaude Code +9 - 0.524score
orchestrate-agents
KSriki
Use to ship a multi-part change by running several agents IN PARALLEL — fan-out across subagents, each isolated in its own git worktree, then verify-and-merge.
SkillBFreshness: BClaude Code +9 - 0.524score
review-pr
KSriki
Use when preparing your own change for review (self-review, commit hygiene, writing the PR description) or when reviewing someone else's pull request for…
SkillBFreshness: BClaude Code +9
Cite this page
handoff skill. Agent Mart. https://agentmart.dev/skills/ksriki-handoff (last updated 2026-10-24)