handoff
Synthesize self-contained implementation prompt(s) that hand a task to a fresh coding agent with an empty context window, decomposing into a phased chain or…
Skill by TheSylvester
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/thesylvester-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: writing, productivity, 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
- Synthesize self-contained implementation prompt(s) that hand a task to a fresh coding agent with an empty context window, decomposing into a phased chain or parallel set when the task is too large for one. Use when the user says "write a handoff prompt", "generate a prompt", "distill this for a fresh agent", or asks to compose a handoff.
- License
- MIT
- Instructions
- 252 lines in SKILL.md · view on GitHub
Files (1)
- SKILL.md
SKILL.md (first lines)
# Handoff Write self-contained prompt(s) to delegate this task to a coding agent with an empty context window. Cover all relevant insights and specifications from the discussion. ## 1. Scope Analysis Before generating, analyze what benefits from shared context: 1. **Identify coupling** — tightly coupled code (same files, shared patterns, mutual dependencies)? Challenge phase boundaries — are they correctly scoped? 2. **Find natural boundaries** — where can work separate cleanly? 3. **Check context fit** — will grouped work fit in one context window (~20 files max)? **Decision:** | Signal | Format | | ------------------------------------------------- | --------------- | | Single focused task, <=3 files, one clear goal | Single prompt | | Multiple distinct file groups or components | Chain | | Discovery + implementation + verification pattern | Chain | | Multiple sequential dependencies | Chain | | Independent work units, no sequencing needed | Parallel | If decomposing, announce: "Task splits into [N] units. Generating [N] prompt files." ## 2. Pre-Generation **Ambiguity resolution** — scan for optionality before generating: | Pattern | Resolution | | -------------------------------------------- | --------------------------------------- | | "Optionally (but recommended)" | **Include** — recommended means do it | | "optional... off by default" | **Skip** — low priority | | "If X, then Y" | Check if X is true in context | | Detailed spec follows the "optional" mention | **Include** — effort was spent on it | | Mentioned in Definition of Done | **Include** — success criterion | Only ask the user if genuinely unresolvable. **Extract for verbatim preservation:** code snippets, protocol specs, exact commands, file paths (absolute), exceptions and nuances. These appear verbatim, never summarized. **Coverage balance** — if one area has significantly more detail than another, flag it before generating. --- ## Prompt Structure (Every Generated Prompt) Phase files may be leaner (reference prior phases, don't repeat). ````markdown <!-- EXECUTION DIRECTIVE: This is a pre-validated implementation prompt. Execute immediately without entering plan mode. --> ## 1. Task One paragraph: what the agent must do and why. ## 2. Context & Constraints - Specifications, goals, success criteria - Constraints, key assumptions, decisions already made ### Prerequisites (chain only) - [ ] Phase [X] completed (if applicable) ### Shared Interfaces (multi-prompt only, CRITICAL) [Exact interface definitions — copy, don't summarize] ## 3. Inputs & Resources ### Files to Create/Modify - Absolute paths ### Files to Reference (Read-Only) - Absolute paths ### Key Code Patterns - Inline code snippets (VERBATIM) ### Build & Test Commands - Exact commands ## 4. Execution Guidelines - Numbered implementation steps - Style and code standards ### Edge Cases - Boundary conditions and handling ## 5. Assumptions | Assumption | Reasoning | |------------|-----------| ## 6. Verification Plan Launch verification sub-agents in parallel: 1. **Test Sub-Agent**: Run `[test command]`. All tests must pass. 2. **Behavioral Sub-Agent**: Prove it works end-to-end. 3. **Code Quality Sub-Agent**: Review for complexity, duplication, security. If any fails, fix and re-run. Retry up to 3 times. ## 7. Definition of Done - [ ] Implementation matches spec - [ ] All verification steps passed - [ ] Assumptions documented (if any) - [ ] No debug code or comments left behind ```` Use sub-agents liberally for parallel work within each prompt. --- ## Where generated prompts go Prompts are written under `.ai-reference/` in the project root — a scratch directory for agent-facing documents that are not part of the shipped codebase. Create it if it does not exist. Suggest adding it to `.gitignore` unless the user wants these prompts committed. If the project already has an established location for this kind of document, use that instead.
From TheSylvester/agentic-coding-tools, MIT.
More from TheSylvester/agentic-coding-tools
- 0.584score
claude-agent
TheSylvester
Run the Claude Code CLI non-interactively as a child agent and get back its final reply plus a session id, with the prompt supplied as arguments, a file, or…
SkillBFreshness: B★ 5Claude Code +9 - 0.584score
codex-agent
TheSylvester
Run the OpenAI Codex CLI non-interactively as a child agent and get back its final reply plus a session id, with the prompt supplied as arguments, a file, or…
SkillBFreshness: B★ 5Claude Code +9 - 0.579score
fusionthink
TheSylvester
Run a multi-vendor adversarial review by dispatching a Claude reviewer and a Codex reviewer over one identical brief, then skeptically verifying every claim…
SkillBFreshness: B★ 5Claude Code +9 - 0.574score
point-by-point
TheSylvester
Explain a topic one point at a time and stop after each point so the user can ask questions or confirm understanding before continuing.
SkillBFreshness: B★ 5Claude Code +9
Signals
- Stars
- 5
- Forks
- 0
- Open issues
- 0
- Last commit (this folder)
- Aug 15, 2026 (2 months 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 TheSylvester
- 0.584score
claude-agent
TheSylvester
Run the Claude Code CLI non-interactively as a child agent and get back its final reply plus a session id, with the prompt supplied as arguments, a file, or…
SkillBFreshness: B★ 5Claude Code +9 - 0.584score
codex-agent
TheSylvester
Run the OpenAI Codex CLI non-interactively as a child agent and get back its final reply plus a session id, with the prompt supplied as arguments, a file, or…
SkillBFreshness: B★ 5Claude Code +9 - 0.579score
fusionthink
TheSylvester
Run a multi-vendor adversarial review by dispatching a Claude reviewer and a Codex reviewer over one identical brief, then skeptically verifying every claim…
SkillBFreshness: B★ 5Claude Code +9 - 0.574score
point-by-point
TheSylvester
Explain a topic one point at a time and stop after each point so the user can ask questions or confirm understanding before continuing.
SkillBFreshness: B★ 5Claude Code +9
Cite this page
handoff skill. Agent Mart. https://agentmart.dev/skills/thesylvester-handoff (last updated 2026-10-24)