Resume from earlier
Skill dhruv-anand-aintech/continue-later-skill/skills/resume-from-earlier
Local handoff skills and CLI for resuming AI coding sessions across Cursor, Claude Code, Codex, Gemini, and OpenCode.
npx -y skills add dhruv-anand-aintech/continue-later-skill --skill resume-from-earlierAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
One thing to look at
- 1 stars1 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.
What its author says it does
Copied from the file, not written here
Resume work from continuation.md or continuation-fast.md. Use when the user runs /resume-from-earlier or asks to pick up from a handoff file—same logic as resume-continuation, shipped as its own skill folder so every platform discovers it by name.
SKILL.md
3.3 KB, as published. Nobody here has run it
Resume From Earlier
Read and summarize project handoff markdown so work can resume quickly without losing context.
This is a separate skill folder from resume-continuation so assistants that only load skills by directory name (not a shared manifest) still pick up the slash-style command /resume-from-earlier. The workflow is identical to resume-continuation.
Files
| File | Typical source |
|---|---|
continuation.md | Full structured handoff (continue-later skill) |
continuation-fast.md | Raw git dump (+ optional transcript excerpts) from continue-later-fast |
Overview
Use this skill when the user says:
/resume-from-earlier(manual attach / slash command)- "resume from earlier"
- "resume from continuation"
- "what was I working on?"
- "continue from last session"
Natural-language variants without a slash are also covered by the resume-continuation skill; either skill may run for the same request.
Workflow
- Resolve which file(s) to read (project root):
- If
continuation.mdexists → treat it as the primary handoff (structured sections). - Else if
continuation-fast.mdexists → read it (often "## Raw Context Dump" / "## Recent user messages"). - If both exist → read
continuation.mdfirst for tasks/state/decisions/gotchas; then skimcontinuation-fast.mdonly when the user asks for the latest git snapshot, raw diff stats, or recent user prompts duplicated there—or when structured sections are missing. - If neither exists → say so and ask whether to create one (continue-later or continue-later-fast).
- If
- Extract only what the user asked for.
- Prefer these sections when present (mostly in structured
continuation.md):## Pending Tasksfor what to do next## Current Statefor working vs broken areas## Key Technical Decisionsto avoid re-litigating choices## Gotchas & Trapsto avoid repeated mistakes## How to Build & Runto restart quickly
Fromcontinuation-fast.md, surface## Raw Context Dump/### Recent user messageswhen relevant.
- Return a concise, actionable resume summary.
Output Format
When asked for a full resume, respond with:
## Resume Snapshot
- **Sources read:** continuation.md / continuation-fast.md / both
- **Current objective:** ...
- **Next 3 tasks:** ...
- **Known blockers:** ...
- **Important decisions:** ...
- **Gotchas:** ...
- **Run commands:** ...
When asked for only one area (for example "show tasks"), return that section only.
Guardrails
- Do not invent project status; only report what exists in the file(s) you read.
- Quote pending tasks verbatim when possible.
- If information is stale or ambiguous, call it out explicitly.
- If the user requests updates, suggest regenerating
continuation.mdwith continue-later, orcontinuation-fast.mdwith continue-later-fast, depending on what they need.