MCP Server for Hidden Markov Model Viterbi Decoder Skill - Pure Python Standard Library implementation of the Viterbi Dynamic Programming Algorithm (Rabiner). Computes the maximum a posteriori (MAP) path of latent states given an observation sequence.
Single tool 'decode_hmm' has a reasonable description (89 chars) and complete input schema with all parameters typed and described. However, the tool name violates verb-noun conventions (should be 'decode_viterbi_hmm' or similar), lacks output schema documentation, and has no error handling guidance. The implementation is mathematically sound but the MCP interface is underdeveloped for production use.
Run Viterbi decoding using log-probabilities to prevent numerical underflow. Returns (optimal_state_sequence, path_log_prob).
Tool name 'decode_hmm' does not follow verb-noun convention. Should start with action verb like 'decode_' or 'compute_'. Current name is ambiguous about what action occurs.
Output schema not documented. Tool returns {optimal_path, log_prob} but LLM has no formal schema to understand field types, enabling downstream tool chaining, or validating responses.
No error handling or recovery guidance. Tool lacks validation of input matrices (e.g., probabilities must sum to 1, no zero probabilities in log-space). Errors return generic 'Unknown tool' message.
Parameter descriptions lack format/constraint details. E.g., 'initial_probs' and 'transitions' are nested dicts but no guidance on required keys, probability ranges, or validation rules.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 43 | 2026-07-28+ | v2 |
No idempotency guarantee documented. Viterbi algorithm is deterministic, but tool description should explicitly state 'idempotent' to signal safe retry behavior to agents.