M-1: dissolve B-08 — canonical modelDisplayName typed at API boundary (N=10 trials) #102

Open
opened 2026-05-12 19:36:50 +00:00 by KofTwentyTwo · 0 comments
KofTwentyTwo commented 2026-05-12 19:36:50 +00:00 (Migrated from github.com)

Goal

Dissolve B-08 by making modelId + modelDisplayName the sole strings referenced in get_self_state MCP tool result text. Harness asserts assistant text never contains 'Opus 4.5' across N>=10 trials when modelId == 'claude-opus-4-7'.

Scope

  • Update MCP get_self_state tool to read SelfState.modelDisplayName exclusively
  • Anti-paraphrase contract enforced in scenario S-SELF-002
  • No free-form model name strings in any prompt or system message

Files

  • MCP tool implementation in App/ or packages/MCP/
  • packages/Harness/Tests/HarnessTests/SelfScenarioTests.swift — S-SELF-002

Acceptance

Test

S-SELF-002 runs N=10 trials in both transport modes.

Parent epic

#5

Source

.planning/architecture/JARVIS-API-MIGRATION-PLAN.md §2.2 (M-1.3)
B-08: #53

## Goal Dissolve B-08 by making modelId + modelDisplayName the sole strings referenced in get_self_state MCP tool result text. Harness asserts assistant text never contains 'Opus 4.5' across N>=10 trials when modelId == 'claude-opus-4-7'. ## Scope - Update MCP `get_self_state` tool to read `SelfState.modelDisplayName` exclusively - Anti-paraphrase contract enforced in scenario `S-SELF-002` - No free-form model name strings in any prompt or system message ## Files - MCP tool implementation in `App/` or `packages/MCP/` - `packages/Harness/Tests/HarnessTests/SelfScenarioTests.swift` — `S-SELF-002` ## Acceptance - [ ] `get_self_state` returns `modelDisplayName` from `SelfState` - [ ] N=10 anti-paraphrase trials: zero occurrences of 'Opus 4.5' when `modelId == 'claude-opus-4-7'` - [ ] Closes #53 ## Test `S-SELF-002` runs N=10 trials in both transport modes. ## Parent epic #5 ## Source `.planning/architecture/JARVIS-API-MIGRATION-PLAN.md` §2.2 (M-1.3) B-08: #53
Sign in to join this conversation.
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: KofTwentyTwo/Jarvis#102