screenkite-clean-cut
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [COMMAND_EXECUTION]: The
scripts/apply_cuts.pyfile usessubprocess.runto interact with the ScreenKite CLI. The call uses an argument list rather than a shell string, which prevents command injection. Evidence: call tosubprocess.run([cli, 'agent', 'tool', 'call', ...]). - [INDIRECT_PROMPT_INJECTION]: The skill handles transcript data. 1. Ingestion points:
compute_all_cuts.pyprocesses a word-level JSON transcript. 2. Boundary markers: None. 3. Capability inventory:subprocess.runinapply_cuts.py. 4. Sanitization: The script only uses timing values and matches text against a fixed list of filler words; transcript text is not interpolated into commands or further prompts. - [EXTERNAL_DOWNLOADS]: The skill uses
uvto manage therequestslibrary. This is a standard and safe dependency management approach.
Audit Metadata