asciinema-converter
Pass
Audited by Gen Agent Trust Hub on Oct 17, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONPROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted terminal recording data (.cast files) which could contain embedded malicious instructions intended to influence the agent when it later reads the converted text files.
- Ingestion points: Reads external .cast files in the file discovery and conversion phases (e.g., in
references/workflow-phases.md). - Boundary markers: No explicit boundary markers or 'ignore' instructions are added to the converted text output to warn downstream processing steps.
- Capability inventory: The skill uses
Bash,Read, andWritetools to perform conversions and manage files. - Sanitization: Uses
sedandtrto strip ANSI escape codes and carriage returns, but this does not neutralize natural language instructions within the terminal output. - [PROMPT_INJECTION]: The
SKILL.mdfile contains a 'Self-Evolving Skill' directive that instructs the agent to 'fix this file immediately' if it encounters issues during use. This creates a self-modification loop where the agent is encouraged to rewrite its own instructions based on runtime experiences, which could be abused by an attacker who can trigger specific error conditions or 'workarounds' to alter the agent's persistent behavior.
Audit Metadata