#53 execute_prompt: schema should advertise cheap/standard/premium tier names as primary model input

[Anonymous]

Originally created by: kumaakh

Problem

The execute_prompt tool schema currently documents model as:

"Model to use (e.g. \"opus\", \"sonnet\", \"haiku\", or full model ID)"

This causes agents reading the schema to translate PM tier names (cheap/standard/premium) into provider-specific model IDs (e.g. premium → opus). This breaks the tier abstraction — the PM skill is designed to never hardcode model names — and triggers security classifier warnings when raw model names appear in dispatches.

Proposed change

• Update the model field description to advertise tier names as the preferred input:

  "Model tier (cheap, standard, premium) or a specific model ID. Prefer tier names — the server resolves them per provider."

• Server continues to accept specific model IDs (for power-user overrides)
• Optionally: log a warning when a raw model ID is passed instead of a tier name, to discourage direct model name usage

Benefit

Agents reading the schema will naturally use tier names, keeping model selection provider-agnostic and avoiding classifier false positives.

[Anonymous]

Originally posted by: kumaakh

Closed based on live testing against v0.1.3_d10302:

• execute_prompt(member_name="focus-dev1", model="cheap", prompt="Reply with exactly one word: hello") → returned "hello" successfully
• Schema model field description: 'Model tier ("cheap", "standard", "premium") or a specific model ID for power users. Prefer tier names — the server resolves them to the correct model per provider.'
• Tier resolution: cheap→haiku-class, standard→sonnet-class, premium→opus-class per provider via modelTiers()

[Anonymous]

Ticket changed by: kumaakh

• status: open --> closed