Guide · Agent skill
How to use the Riffrec Feedback Analysis agent skill
Updated · written with AI from the project's own documentation and checked against our catalogue data
The Riffrec Feedback Analysis skill teaches an AI agent how to read a Riffrec capture, a synchronized screen, voice and event session delivered as a riffrec-*.zip bundle, and turn it into structured evidence for bugs and requirements. It also covers other feedback such as audio, video or notes. This guide explains what the skill does, how to install it, how to ask for it, and what to check before running it on real user data.
Riffrec Feedback AnalysisEveryInc/compound-engineering-plugin · MIT Download the skill Quick start
- 1Install the skill. Download the skill ZIP from this page and unzip it into ~/.claude/skills/ for all projects, or into a project's .claude/skills/ folder for one project.
- 2Confirm Python 3. Make sure a working Python 3 interpreter is on your PATH. The analyzer script will not run without one.
- 3Prepare your input. Have a riffrec-*.zip bundle or another audio, video or notes file ready. Note roughly how long it is and whether it covers one issue or several.
- 4Ask the agent in plain language. Describe the file and the output you want, for example a quick bug report or a requirements extraction.
- 5Let the agent pick the path. The agent routes short single-issue recordings to the quick bug report and longer or multi-issue recordings to extensive analysis.
- 6Review the output before sharing. Check the bug report or findings against the recording, and confirm no sensitive data is in anything you commit or send onward.
What an agent skill is and what this one does
An agent skill is a folder of instructions and optional scripts that an AI agent loads when a request matches the skill's description. This one is the analysis side of Riffrec, a capture tool that records screen, voice and events together.
Once installed, the agent can take a Riffrec bundle or another recording and produce output for downstream work. A short recording (under about 60 seconds, or a single specific issue) yields one evidence-backed bug report. A longer recording, multiple issues, or a workflow walkthrough yields a fuller set of findings and a handoff to a ce-brainstorm step, unless you asked only for extraction.
- Quick bug report: one evidence-backed bug report, with no durable file kept unless you ask for one.
- Extensive analysis: the complete evidence set plus a ce-brainstorm handoff, using the Compound Engineering format.
- Setup help: walks you through installing Riffrec, capturing a session and sharing it, when you have no recording yet.
Install the skill
Download the ZIP for the skill from this page and unzip it. Then place the folder in one of two locations, depending on how you want to use it.
For all your projects, unzip it into ~/.claude/skills/. For a single project, unzip it into that project's .claude/skills/ folder. You can also upload the skill in the Claude apps under Settings → Capabilities → Skills.
- Personal use across projects: ~/.claude/skills/
- One project only: .claude/skills/ inside the project folder
- Claude apps: Settings → Capabilities → Skills, then upload the skill
Check what you need before running it
The skill needs Python 3 for its analysis script. The script also relies on video and audio tooling to produce transcripts and frames; the files reviewed do not specify which tools those are, so check the script's requirements before a first run. Riffrec itself is needed for zip bundles and for the setup path.
The analyzer runs local scripts. The skill's metadata says it does not use the network, but the source also notes that the analysis script processes raw content and may need network or model services for some work. Review that before running it on sensitive material.
- Python 3 on your PATH
- Riffrec capture tool, for zip bundles and setup
- Video and audio tooling used by the script (not specified in the reviewed files)
Trigger the skill with natural requests
You do not need special syntax. Describe what you have and what you want, and the agent should match the request to the skill. Name the artifact, say whether it is a quick issue or a longer walkthrough, and state the output you need.
If you have no recording yet and need help with Riffrec, ask for setup help. The agent should route to the install path rather than analyzing anything.
- "Analyze this riffrec-session.zip and tell me what bug the user hit."
- "Here is a 40-second screen recording of a checkout failure. Write a quick bug report."
- "This is a 15-minute walkthrough of our onboarding. Extract the requirements and hand off to brainstorming."
- "I haven't recorded anything yet. How do I install Riffrec and share a capture?"
- "Just transcribe this meeting recording, no brainstorming."
Understand the output and where files go
By default, raw recordings, audio chunks, zip contents, session dumps and extracted screenshots stay local. In a repository, do not commit raw/ or frames/ folders unless you explicitly decide to and privacy allows it. Text summaries and requirements material can be committed when they contain no sensitive data.
Committed docs should use repo-relative screenshot paths so later readers can open the evidence without absolute local paths. Extensive output follows the Compound Engineering feedback format, with findings that list severity, observed behaviour, expected behaviour, evidence, confidence and requirement candidates.
- Keep raw/ and frames/ out of commits by default
- Use repo-relative paths for screenshots in committed docs
- Expect findings in the Compound Engineering format for longer analyses
What people use it for
- Turning a screen-recorded user complaint into a reproducible bug report for an issue tracker
- Converting a recorded product walkthrough into a requirements document with flows and actors
- Transcribing a meeting or voice note and extracting the action items and requirement candidates
- Getting help installing Riffrec and capturing a first session before any analysis
Frequently asked questions
How do I install the ce-riffrec-feedback-analysis skill?
Download the skill ZIP from this page and unzip it. Put the folder in ~/.claude/skills/ to use it across projects, or in a project's .claude/skills/ folder to use it in one project. You can also upload it in the Claude apps under Settings → Capabilities → Skills.
What do I need to run the Riffrec feedback analysis skill?
You need Python 3 available on your PATH and the Riffrec capture tool for zip bundles. The analysis script also needs video and audio tooling for transcripts and frames, which the reviewed files do not specify. Check the script before first use.
Is the Riffrec feedback analysis skill free?
The skill is released under the MIT licence, as stated on this page. The source does not describe pricing for Riffrec or any model services the agent uses, so check those separately.
Does the skill send my recordings to the internet?
The skill's metadata says it does not use the network. The source also says the analysis script processes raw content and may need network or model services, so review privacy before running it on sensitive or customer data.
What is the difference between quick and extensive analysis?
Quick analysis is for short recordings under about 60 seconds or a single specific issue, and produces one bug report. Extensive analysis is for longer recordings or multiple issues and produces a full evidence set and a ce-brainstorm handoff.
Can I use this skill if I have not recorded anything yet?
Yes, for setup. Ask the agent for help installing Riffrec, capturing a session or sharing a capture. The skill is built around analyzing existing recordings, so it will not produce findings without an input.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
Use this skill if your team already collects screen-recorded feedback and wants it turned into evidence-backed bug reports or requirements. Skip it if you only want general UX advice or if your recordings contain sensitive customer data you are not ready to process locally and review, since the analysis script may use network or model services.
Download the skill