Caveat verdict
bili-summary
Accesses credentials AND writes files
Bilibili video download and Whisper transcription skill that sends transcript text to the Google Gemini API (disclosed, requires user-set GEMINI_API_KEY) for summarization, which is the stated purpose.
Automated static analysis ā not a human review. Caveat flags capabilities, not confirmed intent, and can produce false positives. Disagree with this verdict? Use Dispute below.
What it does
These are capability combinations: each listed behavior occurs in the skill, but Caveat detects co-occurrence ā it does not verify that one flows into another. Read the code to confirm a live chain.
Accesses credentials AND writes files ā may persist stolen credentials locally
LLM02 Ā· LLM06 Ā· ASI03
Permission integrity
file_read+write
credential_access
package_install
Findings (6)
Possible hardcoded credential
SKILL.md Ā· code Ā· API_KEY="your-api-key-here
References sudo ā requests elevated privileges
SKILL.md Ā· code Ā· sudo
Accesses shell history/config
SKILL.md Ā· code Ā· ~/.zshrc
subprocess execution ā runs system commands from Python
scripts/bili-summary.py Ā· prose Ā· downgraded Ā· subprocess.run(
Python urllib.request ā network access
scripts/bili-summary.py Ā· prose Ā· downgraded Ā· urllib.request
Python os.environ.get ā reads environment variable
scripts/bili-summary.py Ā· prose Ā· downgraded Ā· os.environ.get(
Why the tier is capped
Execution sink present in raw bytes (Hard Floor: class D/F). Final tier capped at Caution ā cannot be lifted by any downgrade, example-payload opt-in, or allowlist.
Permissions & capabilities
No declared permissions ā minimal attack surface.
package_installfile_writecredential_access Thanks ā recorded.