timeline-asciinema
Pass
Audited by Gen Agent Trust Hub on Sep 4, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes potentially untrusted terminal recording data which could contain malicious instructions designed to influence the agent's parsing logic.
- Ingestion points: Processes .cast and .json files from the input/terminallogs/ directory.
- Boundary markers: While it mentions stripping prompts and ANSI sequences, it lacks explicit delimiters or instructions to ignore embedded natural language commands within the recordings.
- Capability inventory: Primarily focused on reading logs and writing structured data to output/terminal_entries.json.
- Sanitization: Includes steps to strip ANSI escape sequences and control characters before performing command inference.
Audit Metadata