requirements.txt

# markitdown-skill dependencies
#
# Bounded (upper-capped) rather than fully pinned, so the skill keeps working
# with patch releases while a breaking major/minor cannot slip in silently.
# For byte-reproducible installs, replace the range with the exact version you
# verified locally (see `pip show markitdown`) — ClawHub/SkillHub reviewers
# specifically look for a deliberate, documented pinning strategy.
markitdown[all]>=0.1.3,<1.0

# Optional, only needed for `--llm-model` image descriptions / document analysis.
# Disabled by default and never used without explicit user consent.
# openai>=1.0,<2.0

# System (non-pip) dependencies, install via your package manager:
#   ffmpeg    - audio/video transcription        (apt: ffmpeg / brew: ffmpeg)
#   chromium  - SPA render fallback on Linux     (apt: chromium / brew: chromium)
#   exiftool  - optional image EXIF metadata     (apt: libimage-exiftool-perl)