analyze-screen-feedback
v1.0.0by andydrewie · author: Andrew Fai
Ground narrated macOS screen-recording feedback in local word timestamps, visible interface evidence, cursor motion, and explicit uncertainty.
What analyze-screen-feedback does, in the publisher's words
analyze-screen-feedback is a local-first Codex skill for grounding narrated screen-recording feedback in visible interface evidence. It aligns word timestamps, low-confidence speech, cursor and screen motion, selected keyframes, and uncertainty so references such as "this," "here," numbers, pointing, and circling can be resolved without guessing.
- Inspect media streams and local processing capabilities before doing work.
- Extract mono audio without modifying the source recording.
- Produce word-level timestamps with a pinned MLX Whisper configuration.
- Find motion windows and select readable timestamped keyframes.
- Emit a compact evidence bundle with transcript, motion observations, frames, provenance, and report rows.
- Guide Codex to distinguish observation from inference, combine audio and visual confidence, and preserve unresolved alternatives.
The script intentionally does not name interface targets by itself. Codex must confirm candidate objects from readable frames and use the same evidence contract if native multimodal capabilities eventually replace local preprocessing.
From the project README, punctuation lightly normalized · Open the README on GitHub
- Skills
- 1
- MCP servers
- 0
- Stars
- 0
- License
- MIT
- Repo created
- 2026-08-06
- Last pushed
- 2026-08-07
- Publisher type
- User
- Version
- 1.0.0
Links
Skills · 1
- analyze-screen-feedback
- Analyze narrated macOS screen recordings by aligning local word-level transcription, cursor and screen motion, timestamped frames, visible interface objects, and uncertainty. Use for screen-recording feedback, usability walkthroughs, bug narrations, design critiques, or any request to resolve spoke…
Descriptions come from the frontmatter of each SKILL.md, punctuation lightly normalized.
Category
Agent Tooling. Assigned by this directory. Agent Plugins 1.0.0 has no category field, so no manifest declares one.
Keywords
screen-recording · usability · multimodal · macos · feedback
Related by publisher and keywords
Transform verbose instructions into concise prompts and canonical technical terms while preserving every stated requirement.
Add decision-useful quantitative structure for scale, comparison, likelihood, economics, uncertainty, and consequential empirical claims without numerical theater or false precision.
UX and UI design skills including Refactoring UI, iOS HIG design, usability heuristics, habit-forming product design, behavior design for retention, web typography, award-winning top-design, Steve Jobs-style design reviews, and Don Norman's foundational design principles
Expert guidance on Swift Concurrency best practices, patterns, and implementation. Covers async/await, actors, tasks, Sendable conformance, data race prevention, and Swift 6 migration strategies.
Secure, profile-scoped access to a local Windows or macOS workstation through WorkForge.
Visual markup on a running dev server, handed to the agent as one locked batch. Registers Tailr's MCP server and the operating rules for the review loop.
[](https://agentpluginsdirectory.com/plugins/analyze-screen-feedback)✓ Verified . We fetched the manifest from GitHub and checked it against the official Agent Plugins 1.0.0 schema at agent-plugins.org.