feat(extract): write the omitted-file list where a user can read it

[L4XB]

Part 3 of #3574.

The incomplete-extraction warning names five files and counts the rest:

[graphify extract] semantic extraction is incomplete: 73 dispatched file(s)
produced no nodes and 0 came back truncated or hollow. ...

and the reporter had to diff the graph against a filesystem walk to find out which 73. The list was never missing: llm.py already puts every one of them on the merged result as uncovered_files, and cli.py reads that only for its length and then drops it.

The change

cli.py writes it to graphify-out/.graphify_uncovered_files.json, next to the other dot-prefixed artifacts, and the warning ends with the path so the next step is obvious rather than archaeological.

Two boundaries, each with a cell:

• A clean run clears it. A list left over from an earlier run describes files that are covered now, which is worse than no list at all, so an empty result unlinks the file.
• It never costs the extraction. An output directory that refuses the write loses the diagnostic and nothing else, the same stance backup_if_protected takes.

Test

Three cells in tests/test_hollow_chunks_arm_shrink_guard.py, which is where the _omitted_files path is already exercised end to end through mainmod.main(). Two of them drive the real CLI rather than the helper: one asserts the written list and that the warning names its path, the other seeds a stale list and asserts a covered run removes it.

All three fail on v8 (ImportError, the API is new). The suite is green at 8, and the related slices at 350 passed, 6 skipped. The one failure in that run, test_ollama_retry_cap.py::test_api_timeout_is_passed_to_client, also fails on unmodified v8 here.

Mutation-checked: four seeded defects — never writing, keeping a stale list, dropping the path from the message, and letting a refused write abort the run — each caught by a named cell, none surviving.

Parts 1 and 2 are separate. Part 1 is #3576. Part 2, the AST cache pinning ids to the extraction root, I could not reproduce on v8; the details are in a comment on the issue.

[graphify-labs]

Graphify reviewed this change.

Looks safe to merge — no coupling regressions and no blocking issues, checked against the code graph (not a self-assessment).

Formal verification. No changes could be formally verified in this run.

---

Graphify review — findings

Adds a .graphify_uncovered_files.json sidecar written into the graphify-out directory during semantic extraction that records the full sorted list of dispatched files that produced no nodes, so the list is recoverable instead of only counted in the incomplete-extraction warning. The warning now appends the path to that file when it was written, an empty list deletes any stale sidecar from an earlier run, and _write_uncovered_files swallows OSError so a read-only output directory drops the diagnostic without failing the extraction.

No blocking issues surfaced. 2 lower-confidence candidates did not survive cross-model review.

Analysis details — impact, health, verification

Impact & health

Graphify review

Impact — 294 functions depend on the 81 functions this change touches.

Health — this change adds coupling hotspots:

• new: dispatch_command() — 2 callers, 125 callees
• new: _stale_graph_sources() — 7 callers, 6 callees
• new: _run_hook_guard() — 4 callers, 8 callees
• new: test_poisoned_manifest_is_healed() — 0 callers, 6 callees

Verification — 294 functions in the blast radius were not formally verified this run (proofs are advisory here).

Gate & verification

graphify gate

PASS — objectively clean (no health regressions, tests not run — proofs not run this pass (advisory)). Grounded, not self-assessed.

Advisory (not blocking):

• verification_scope: 237 function(s) in the blast radius were not formally verified this run

Test selection

Test selection

24 of 275 test file(s) selected (9%) via static blast radius.

• tests/test_affected_cli.py — impact
• tests/test_agents_platform.py — impact
• tests/test_codebuddy.py — impact
• tests/test_devin.py — impact
• tests/test_explain_cli.py — impact
• tests/test_extract_cli.py — impact
• tests/test_global_add_tag_inference.py — impact
• tests/test_god_nodes_cli.py — impact
• tests/test_hollow_chunks_arm_shrink_guard.py — impact, changed-test
• tests/test_hook_guard_token_match.py — impact
• tests/test_hook_out_of_project_paths.py — impact
• tests/test_hook_strict.py — impact
• tests/test_incomplete_build_guard.py — impact
• tests/test_install.py — impact
• tests/test_install_references.py — impact
• tests/test_merge_chunks_validation.py — impact
• tests/test_multigraph_diagnostics.py — impact
• tests/test_no_dedup_flag.py — impact
• tests/test_partial_cache.py — impact
• tests/test_path_cli.py — impact
• tests/test_query_cli.py — impact
• tests/test_query_induced_edges.py — impact
• tests/test_stale_prune.py — impact
• tests/test_unverified_semantic_shrink.py — impact

Selection is safe under the controlled-regression assumption; always-run tests + a periodic full run are the backstops. Advisory — it never changes the check verdict.

Formal verification

Could not verify: Could not verify dispatch\_command.

The verifier did not have enough to check dispatch\_command, so it is saying so rather than guessing. No false assurance is the whole point.

Guarantee: No guarantee either way, this is an honest abstention, not a pass.

Note: Reason: not verifiable: all 23 sampled inputs raised on both versions — the function never executed, so 'no divergence' would be vacuous (mostly SystemExit — names the real obstacle, not a sampling gap)

· 4 more finding(s) on lines outside this diff (see the check run).