Skip to content

Skill Recorder v0.8.0

Latest
Latest

Choose a tag to compare

View all tags
@GiorgioUghini GiorgioUghini released this 02 Nov 16:49
v0.8.0
505c97a
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

Source-only release

Skill Recorder 0.8.0 is published from commit 505c97ae47be5642c0e1b44ebb8d1c1530c298fa.

This release is source-only. No installers or portable binaries are attached. GitHub's generated archives contain the tagged repository source only.

Changes

  • Import an existing recording: use the smaller folder button beside Record to select a local MP4 or WebM, such as a downloaded Teams demonstration.
  • Check actual video and audio decoding before proceeding. Unsupported codecs and invalid media are rejected with an explanation; no converter, standalone FFmpeg installation, or new application dependency is required.
  • Add optional details to explain the task, supplement the demonstration, or suggest where it appears. Leaving details empty is supported. Timestamps guide the AI; they do not trim the recording.
  • Transcribe audio locally with the existing Whisper model, then reconstruct activities from the video, speech, and user details. Silent videos and videos without audio are supported; transcription failures stop analysis so they can be retried.
  • Reuse the existing activity-review and skill-building flow: edit, reorder, add or remove activities, then create, review, and install/export a skill.
  • Preserve live screen recording, its settings, and its event-first analysis behavior.
  • Keep the original recording untouched. Store a local session copy with cancellation, retry, and privacy protection for outgoing frames and text.

Compatibility and limitations

  • Import limits: 512 MiB, 15 minutes, and up to 3840 x 2160 pixels (or an equivalent pixel count). Audio must be mono or stereo at no more than 48 kHz. One video track and at most one audio track are supported.
  • MP4/WebM is a container, not a guarantee of codec support: the shipped Chromium runtime must be able to decode the selected file. Chromium uses the already-shipped, reviewed codec library; no separate FFmpeg executable is added.
  • CPU-based transcription can take several minutes for longer recordings. The current status does not show per-chunk progress, and canceling during inference may wait for a processing boundary.
  • A conservative synthetic 512 MiB / 15-minute stereo stress case with Whisper resident peaked at approximately 4.6 GB (4.25 GiB) across Electron processes on Windows ARM64. This is not a universal memory ceiling and excludes other apps and the external Copilot CLI process. The import decoder has a 2 GiB working-set guard.
  • Imported video lacks the OS events captured during live recording. Hidden clipboard contents, unreadable commands, and actions neither shown nor spoken cannot be recovered reliably; review the inferred activities before creating a skill.
  • Selecting a file does not upload the recording. Choosing Analyze sends selected frames, transcribed speech, and additional details to GitHub Copilot through the existing protection pipeline. Only import recordings you are authorized to use; protection is a safety net, not a guarantee.
  • Dependencies and installer scripts are unchanged. Re-run the commit-pinned source installer below to update. Locally assembled source builds are not redistributable release binaries.

Validation

  • Local validation: 267 existing regression tests and 8 import tests, production build/typecheck, eval typecheck, lockfile portability, and a 285-package license inventory.
  • Reviewed Electron Windows ARM64 runtime checks covered MP4/WebM import, absent/silent audio, unsupported audio/video rejection, seek/crop, cancellation, original-file preservation, exact 512 MiB input, and one-byte-over rejection.
  • A real imported recording completed local transcription and generated an activity analysis.
  • Windows x64 and ARM64 and macOS and Ubuntu passed on the exact release commit, covering commit-pinned source installation, tests, builds, compliance, and applicable packaging/architecture checks.
  • Synthetic media runtime checks were performed locally on Windows ARM64; the cross-platform workflows are not a claim that every codec or Teams export was manually exercised on every platform.

Installer script SHA-256

These hashes were calculated from the exact release commit's Git blobs.

7678f7f8d38d600f2fb732f95afeadc8aa4f9b9802989d26e2f932970795770c  install.ps1
bd8b5200a2c140c5a5ed9299d565ac72cee5420181774008977eaf75f210db1d  install.sh

Commit-pinned installation

Windows 11 x64 or ARM64

$commit="505c97ae47be5642c0e1b44ebb8d1c1530c298fa"; $env:SKILL_RECORDER_COMMIT=$commit; irm "https://raw.githubusercontent.com/microsoft/skill-recorder/$commit/install.ps1" | iex

macOS or Ubuntu

commit="505c97ae47be5642c0e1b44ebb8d1c1530c298fa"; curl -fsSL "https://raw.githubusercontent.com/microsoft/skill-recorder/$commit/install.sh" | SKILL_RECORDER_COMMIT="$commit" bash

For script inspection and hash verification before execution, use the inspect-first instructions.

Full changelog: v0.7.0...v0.8.0

Assets 2
Loading