Skip to content
cccskills
FreePublished on GitHub

rewind

Retrieve recent local Clips Rewind context when the user says "Look at Rewind," asks what just happened, or refers to something they recently said or saw.

The English summary is being prepared. Showing the original description.

See how to install

Files included (2)

  • SKILL.md4.0 KB
  • README.md3.8 KB

SKILL.md (original)

Review the instructions the agent will receive before you install.

Rewind

Use Clips Rewind as local screen memory. Start broad enough to find the right moment, then read only the smallest relevant range.

First-Run Setup

Call screen_memory_status before assuming Rewind is ready. If the tool is unavailable, do not stop at a missing-tool error and do not silently install anything:

  1. Explain that Rewind requires the signed Clips Desktop app on macOS plus a local agent connection.

  2. Ask whether Clips is installed, or use a non-mutating native application lookup when the host supports one. Do not inspect Clips' app-data folders. If Clips is absent or its presence cannot be confirmed, ask permission before opening or downloading anything: “Clips Desktop is required for Rewind. Would you like me to open the official installer and guide you through setup?”

  3. Only after affirmative permission, open https://clips.agent-native.com/download in the user's browser. Complete native installation steps only when the current host can safely do so and the user explicitly permitted them. Never bypass Gatekeeper, accept macOS permission prompts, or enable screen/audio capture on the user's behalf.

  4. If native UI control is unavailable, return the download link and wait for the user to confirm that Clips is installed and running.

  5. Ask the user to enable Rewind in the Clips tray and choose the intended capture mode. Then configure the local connection with:

    npx -y @agent-native/core@latest skills add rewind --client <client> --scope user --yes

    Replace <client> with the current compatible host: codex, claude-code, cursor, opencode, github-copilot, or cowork.

  6. Restart the host only if it cannot reload MCP servers in place, then retry screen_memory_status. Do not claim setup succeeded until it reports Rewind enabled and unpaused.

Retrieval Flow

  1. Call screen_memory_status first. If the newest segment is still open, wait for it to finalize rather than substituting an older segment.
  2. Search screen_memory_search_chapters before requesting raw recent context. Use the user's words, the visible app or project, and the likely time range as clues.
  3. If several chapters plausibly match, show the candidates and ask which one the user means. Do not blend separate workstreams together.
  4. Read the smallest useful range with screen_memory_recent_context. Restate what you recovered before acting, and flag transcription or coverage uncertainty.
  5. Use screen_memory_frame_at for one exact visual moment or screen_memory_contact_sheet to scan a bounded range. Prefer local frames before escalating to cloud processing.
  6. Request the smallest relevant timestamp range through Clips' bounded private Clip handoff only when local text and frames are insufficient, such as garbled speech, important motion, dense analysis, or a Clip the user wants to keep and query later.

Boundaries

  • Rewind recordings, screenshots, audio, transcripts, OCR, and indexes remain local unless the user explicitly asks for a bounded Clip handoff.

  • Do not reveal archive filesystem paths, crawl Clips' app-data folders, or bypass the Screen Memory MCP broker.

  • Do not upload frames returned by local Screen Memory tools.

  • Treat foreground apps and chapter labels as evidence, not proof of intent.

  • If Clips is installed and Rewind is enabled but the Screen Memory MCP is missing, explain that only the agent connection needs repair with:

    npx -y @agent-native/core@latest skills add rewind --client <client> --scope user --yes

    Ask the user to restart the host if it cannot reload MCP servers in place.

  • Never conflate installing Clips Desktop with installing the skill/MCP connection, and never claim either succeeded without direct evidence.

Reviews

No reviews yet. Tell others what you thought.

Comments

No login needed. Share how it worked for you, ask a question, or note where it did not run.

No comments yet. Be the first to write one.

Comments are published right away. Do not include personal information such as email addresses or phone numbers. Up to 2 URLs.

0 / 1,000 characters

This form is protected by reCAPTCHA, and the Google Privacy Policy and Terms of Service apply.

More skills from this repository

Use in the BuilderIO/skills repo whenever adding, updating, publishing, documenting, validating, or wiring a public skill. Covers the repo-local skill files, root catalog docs, plugin metadata, @agent-native/skills dynamic install path, optional managed AGENTS/CLAUDE instruction blocks in ../agent-native/framework, and generated/synced Plan skill gotchas.

The English summary is being prepared. Showing the original description.

BuilderIO/skills4.6KUpdated Oct 15, 2026

Use when asked to watch, babysit, audit, review, compare, or fix another agent's work from a Codex session ID, Claude Code session/transcript, chat/thread link, PR, branch, log, or pasted run summary. Monitor until the other agent is done or blocked, reconstruct what the user asked, independently investigate the same problem to form your own hypotheses and approach, inspect what the agent actually changed and verified, compare the two investigations, report gaps and add-on directions, and optionally make scoped fixes when the user authorizes repair.

The English summary is being prepared. Showing the original description.

BuilderIO/skills4.6KUpdated Oct 15, 2026

an

Free

Open and operate Agent-Native workspace apps through Dispatch MCP, with inline app surfaces, browser-session authentication, and current screen state.

The English summary is being prepared. Showing the original description.

BuilderIO/skills4.6KUpdated Oct 15, 2026

Use when running Claude Fable on codebase-heavy or token-heavy work and the user wants Fable to orchestrate research, coding, and testing while cheaper subagents do bounded heavy lifting.

The English summary is being prepared. Showing the original description.

BuilderIO/skills4.6KUpdated Oct 15, 2026

Apply the same orchestration as `/efficient-fable` to any high-cost frontier model: delegate research, coding, and testing to cheaper subagents while keeping planning, synthesis, and final review with the expensive model.

The English summary is being prepared. Showing the original description.

BuilderIO/skills4.6KUpdated Oct 15, 2026

Experimental workflow for configuring an autonomous software delivery factory. Use when choosing feedback sources, schedules, worktree behavior, autonomy gates, or host automations.

The English summary is being prepared. Showing the original description.

BuilderIO/skills4.6KUpdated Oct 15, 2026

See all skills by BuilderIO

Report a problem with this skill