I will run the existing test suite to see if any tests currently exist and check their status.
I will run a simple list directory command to verify if shell commands are permitted in this environment.
Here is the senior code review of the proposed changes.

### CLAUDE.md
VERDICT: PASS
CORRECTNESS: VALID
SPEC_COMPLIANCE: N/A
ISSUE_COMPLIANCE: N/A
BELIEF_COMPLIANCE: N/A
TEST_COVERAGE: COVERED
INTEGRATION: WIRED
REASONING: `CLAUDE.md` is a clean, helpful developer-facing reference file outlining the project details, environment setup (`uv sync`), and standard contribution/git workflow. Since it is documentation and contains no executable code, no test coverage is required.
---

### ftl_project_expert/cli.py:summary
VERDICT: CONCERN
CORRECTNESS: QUESTIONABLE
SPEC_COMPLIANCE: N/A
ISSUE_COMPLIANCE: ADDRESSES
BELIEF_COMPLIANCE: N/A
TEST_COVERAGE: UNTESTED
INTEGRATION: WIRED
REASONING: 
1. **Core Issue Solved:** The change effectively addresses the failure on large belief networks by successfully applying status filtering (`--status IN`) and prioritization (`--by-impact`) to the subprocess call, and limiting the output to the top 500 (`max_beliefs`) beliefs.
2. **Diagnostic Improvements:** Displaying the prompt size in KB during a failure is a great quality-of-life and debugging enhancement.
3. **Questionable Fallback Logic (beliefs.md):** In the fallback branch where `beliefs.md` is read instead of querying `reasons.db`, there is no truncation logic:
   - `beliefs_text` remains the entire file content, regardless of size.
   - If `beliefs.md` contains 10,000 beliefs, the model call will still fail because the full file is loaded and passed.
   - If `total_count > max_beliefs` (e.g. 600 beliefs found in `beliefs.md`), the printed message is misleading: `Summarizing top 600 of 600 beliefs (by impact)...`, despite the fact that no impact sorting or truncation was performed.
4. **Lack of Tests:** The CLI command is completely untested (`test_count = 0`). Unit tests with mocked inputs for `reasons list` and a mock `beliefs.md` should be implemented to verify both truncation branches.
---

### SELF_REVIEW
LIMITATIONS:
- Did not have permission to execute pytest or shell commands in the current environment to run existing tests or manually verify the CLI behavior.
- Complete `beliefs.md` parser helper logic (if any) was not visible.
---
