ARCHIVED [experiment over]: An AI agent built to do Ralph loops - plan mode for planning and ralph mode for implementing.

fix: Phase 1d re-review issues — restore test helpers, fix assertions, clean imports master

- Restore original test helpers in common/mod.rs that were destroyed by bug-fixer's MockGraphStore consolidation (C1) - Fix resolve_agents_md ordering: remove erroneous .rev() so results are closest-to-file first as tests expect - Replace vacuous assert!(true) in token budget warning test with real assertion checking wrap-up message injection in recorded LLM calls (I1) - Fix mock token counts test setup: set_token_counts is global, not per-response, so remove overwriting second call - Remove unused imports from agent_runtime_test.rs and agent_types_test.rs (M1) - Add #[allow(dead_code)] on AgentRuntime.profile field stored for future multi-agent orchestration (M2) Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>


+219 -29
5 changed files