fit-hidden-markov-model
Pass
Audited by Gen Agent Trust Hub on Mar 18, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill contains purely informational and procedural content regarding the implementation of statistical algorithms (Baum-Welch, Viterbi). No malicious behavior or security risks were identified.
- [PROMPT_INJECTION]: The instructions do not contain any patterns typical of prompt injection, such as attempts to override system prompts, bypass safety guidelines, or ignore previous instructions.
- [DATA_EXFILTRATION]: There are no commands or instructions that access sensitive local files (e.g., SSH keys, cloud credentials) or perform unauthorized network communication.
- [REMOTE_CODE_EXECUTION]: The skill does not reference any external scripts, remote URLs for execution, or unverifiable third-party dependencies.
- [COMMAND_EXECUTION]: While 'Bash' is listed as an allowed tool, the procedure is restricted to describing the mathematical steps for the algorithm and does not instruct the agent to execute dangerous system commands.
- [DATA_EXPOSURE]: The skill operates on 'observations' data provided by the user but does not expose any internal or sensitive information. It includes a basic surface for indirect prompt injection via the 'observations' input, but since the instructions are purely mathematical, the risk is minimal.
Audit Metadata