Handoff

Skill Adit-Jain-srm/skill-forge/skills/handoff

Compact the current conversation into a handoff document so another agent or future session can continue the work without losing context. Use when ending a session, switching tasks, the context is getting long, or user says handoff, wrap up, save progress, or "continue this later".From its SKILL.md

Install

npx -y skills add Adit-Jain-srm/skill-forge --skill handoff

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

2 stars2 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

2.7 KB, 608 tokens by cl100k_base, as published. Nobody here has run it

Overview

Create a structured document capturing everything needed to continue work in a fresh session — in under 50 lines.

Process

Write a handoff document capturing everything needed to continue this work.

What to Include

1. GOAL — what we're building/fixing (one sentence)
2. CURRENT STATE — what's done, what's in progress, what's blocked
3. KEY DECISIONS MADE — choices that shouldn't be re-litigated
4. OPEN QUESTIONS — unresolved decisions awaiting input
5. NEXT STEPS — exact actions to take (not vague "continue working")
6. FILES TOUCHED — paths modified this session
7. SUGGESTED SKILLS — which skills the next session should invoke

What NOT to Include

Full code (reference by file path instead)
Conversation history (this is a SUMMARY not a transcript)
Sensitive info (API keys, passwords, PII)
Resolved dead ends (only include what's RELEVANT going forward)

Where to Save

Save to the OS temp directory (not the workspace):

Windows: $env:TEMP/handoff-{project}-{date}.md
macOS/Linux: /tmp/handoff-{project}-{date}.md

Tell the user the path so they can reference it next session.

Common Mistakes

Including full code instead of file paths (bloats the handoff, makes it unreadable)
Forgetting open questions (next session re-litigates decisions)
Writing vague next steps like "continue working" (be SPECIFIC: which file, which function, which test)
Saving to the workspace (pollutes the project — use temp directory)

Format

Keep it SHORT. Under 50 lines. A good handoff is one page a new agent can scan in 10 seconds and know exactly what to do.

Example

## Handoff — skill-forge v1.2.0

**Goal:** Add TF-IDF semantic routing to replace token-overlap matching.

**State:** Route-task.js rewritten and passing. 98 tests green. Not yet pushed.

**Decisions Made:**
- TF-IDF + cosine similarity (not embedding API — zero dependencies)
- Compound routing: combine top 3-5 skills when overlap < 40%
- Indexed learnings surfaced alongside skill matches

**Open Questions:**
- Should RL weights be per-skill or per-category? (currently per-category)

**Next Steps:**
1. Run `node tests/run-tests.js` — confirm all pass
2. Push with `git -c http.version=HTTP/1.1 push origin main`
3. Update SKILL.md Phase 4 to reference new routing

**Files:** scripts/route-task.js, memory/indexed-learnings.json, tests/run-tests.js
**Skills for next session:** /prove-it, /self-review

What ships with it

Read from the repository

Just SKILL.md. No reference files, no scripts.

Gives 0 of the 12 instructions most agent orchestration skills give in 608 tokens

Counted across 848 of the 1,300 authors here whose files we hold, read 2026-09-06

Dispatch one agent per independent problem domainin 56 of 848, across 42 files
Run full test suite after integrationin 55 of 848, across 42 files
Verify fixes do not conflictin 40 of 848, across 32 files
Review each summary when agents returnin 40 of 848, across 31 files
Write a handoff document summarising the current conversationin 30 of 848, across 25 files
Reference existing artifacts by path or URLin 26 of 848, across 24 files
Give each agent a specific scopein 19 of 848, across 10 files
Give each agent a clear goalin 19 of 848, across 10 files
Include a suggested skills section in the documentin 18 of 848, across 16 files
Tailor the doc to the user argumentsin 18 of 848, across 15 files
Issue all subagent dispatches in the same responsein 17 of 848, across 11 files
Use git worktrees for isolationin 17 of 848, across 8 files

Said here and by no other author read

Keep the document under 50 lines

Grouped from the skills themselves: near-identical wordings counted once, and counted by distinct author, so one author publishing three of these counts once. Length counted with cl100k_base; the agent that loads this file may tokenize it differently.