菜单
Use when saving or restoring agent context across sessions -- the user asks to save state, create a handoff, or pause, or the context window nears capacity; or to resume with load handoff, resume from, or continue where we left off.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
| name | session-handoff |
| description | Use when saving or restoring agent context across sessions -- the user asks to save state, create a handoff, or pause, or the context window nears capacity; or to resume with load handoff, resume from, or continue where we left off. |
| tags | ["meta"] |
| tier | core |
Creates handoff documents that enable fresh AI agents to continue work without ramp-up time. Solves the long-running agent context exhaustion problem.
Determine which mode applies:
Creating a handoff? User wants to save current state, pause work, or context is getting full.
Resuming from a handoff? User wants to continue previous work, load context, or mentions an existing handoff.
Proactive suggestion? After substantial work (5+ file edits, complex debugging, major decisions), suggest:
"We've made significant progress. Consider creating a handoff document to preserve this context for future sessions. Say 'create handoff' when ready."
Run the smart scaffold script to create a pre-filled handoff document:
python scripts/create_handoff.py [task-slug]
Example: python scripts/create_handoff.py implementing-user-auth
For continuation handoffs (linking to previous work):
python scripts/create_handoff.py "auth-part-2" --continues-from 2024-01-15-auth.md
The script will:
handoffs/ directory in the tool's config directory (e.g., .claude/handoffs/) if neededOpen the generated file and fill in all [TODO: ...] sections. Prioritize these sections:
Use the template structure in references/handoff-template.md for guidance.
Run the validation script to check completeness and security:
python scripts/validate_handoff.py <handoff-file>
The validator checks:
[TODO: ...] placeholders remainingDo not finalize a handoff with secrets detected or score below 70.
Report to user:
List handoffs in the current project:
python scripts/list_handoffs.py
This shows all handoffs with dates, titles, and completion status.
Before loading, check how current the handoff is:
python scripts/check_staleness.py <handoff-file>
Staleness levels:
The script checks:
Read the relevant handoff document completely before taking any action.
If handoff is part of a chain (has "Continues from" link), also read the linked previous handoff for full context.
Follow the checklist in references/resume-checklist.md:
Start with "Immediate Next Steps" item #1 from the handoff document.
Reference these sections as you work:
As you work:
--continues-from to chain themFor long-running projects, chain handoffs together to maintain context lineage:
handoff-1.md (initial work)
↓
handoff-2.md --continues-from handoff-1.md
↓
handoff-3.md --continues-from handoff-2.md
Each handoff in the chain:
When resuming from a chain, read the most recent handoff first, then reference predecessors as needed.
Handoffs are stored in the tool's config directory, e.g., .claude/handoffs/ for Claude Code.
Naming convention: YYYY-MM-DD-HHMMSS-[slug].md
Example: 2024-01-15-143022-implementing-auth.md
| Script | Purpose |
|---|---|
create_handoff.py [slug] [--continues-from <file>] | Generate new handoff with smart scaffolding |
list_handoffs.py [path] | List available handoffs in a project |
validate_handoff.py <file> | Check completeness, quality, and security |
check_staleness.py <file> | Assess if handoff context is still current |
Use when creating, editing, reviewing, or rendering software architecture or flow diagrams — C4 diagrams (system context, container, component, deployment, dynamic), sequence diagrams, request-flow visuals, or any Mermaid authoring where the rendered output must be readable. Also use when a Mermaid diagram renders with overlapping labels, parse errors, or a single-column layout, or when the user asks to "diagram the architecture", "draw the request flow", "map the sequences", or wants LucidChart-quality output from Mermaid.
Use when about to create a feature, build a component, add functionality, modify behavior, or start any new work — or when the user asks how should we, what approach, which way, explore options, design X, or spec this out. Triggers before any code, scaffold, or implementation skill runs.
Use when one request spans multiple domains (e.g. code + tests + docs + security + diagrams), has independent workstreams that could run concurrently, or is too large for a single agent in one pass — symptoms include "do all of X, Y, and Z", a multi-part review, or a task that would obviously thrash one agent's context
Use when defining product requirements, writing or updating a PRD, spec'ing out a feature, or deciding what to build — or when the user says write requirements, create a PRD, define the product, what should we build, or hands over stakeholder/engineering feedback to fold into an existing PRD. Especially for non-technical users translating a conversation into structured requirements.
Use when you want brutal, worst-case review of code resilience -- failure modes, race conditions, blast radius, and edge cases optimistic reviewers miss.
Use when reviewing code or a PR for high-confidence, real-exploitation security vulnerabilities (>80% confidence threshold, low false positives).