fix(store): guard store_findings against None engagement_score on update path

[nicolefinateri]

What

store_findings raises TypeError when a re-sighted finding carries engagement_score: None.

Why

On the update branch the incoming score is read with a defaulted .get, then compared:

new_engagement = f.get("engagement_score", 0)            # store.py:427
...
max(new_engagement, existing["engagement_score"] or 0)   # store.py:430

dict.get(key, 0) only substitutes 0 when the key is absent. A key that is present with value None passes straight through, so new_engagement becomes None and max(None, <float>) raises:

TypeError: '>' not supported between instances of 'float' and 'NoneType'

Note the asymmetry: the existing side on line 430 is already guarded with or 0, but the new_engagement side is not. engagement_score is declared float | None in lib/schema.py, and lib/fusion.py can re-introduce None via metadata.get("engagement_score"), so None is a first-class runtime value for this field.

Scope (honest framing)

This is boundary hardening, not a fix for a crash reachable through the normal pipeline. The two producers that feed store_findings via findings_from_report — finding_from_candidate (candidate.engagement or 0) and the supplement dict (item.engagement_score or 0.0) — already coerce None → 0 before the dict is built. But store_findings(List[Dict[str, Any]]) is a public function that accepts arbitrary dicts, so any other or future caller passing a raw finding with engagement_score: None hits the crash. This makes the boundary null-safe and symmetric with line 430.

Fix

-            new_engagement = f.get("engagement_score", 0)
+            new_engagement = f.get("engagement_score") or 0

Test

Adds test_store_findings_none_engagement_on_update_does_not_crash to tests/test_store.py: it stores a finding with a real score, re-sights the same URL with engagement_score: None, and asserts no crash plus that max(None→0, existing) preserves the higher real score. The test reproduces the TypeError on main and passes with this change.

Verified locally (Python 3.12.13, pytest 9.1.0, deps from uv.lock):

• tests/test_store.py — 28 passed (the 27 existing tests + this one)
• store-consumer blast radius (test_store, test_internals_v3, test_watchlist_commands, test_fusion_v3, test_signals_v3, test_briefing_v3, test_schema_v3) — 179 passed, 0 failed

[greptile-apps]

Greptile Summary

This PR makes store_findings tolerate None engagement scores on re-sighted findings. The main changes are:

• Normalizes missing or falsy incoming engagement_score values to 0 before update comparisons.
• Adds a regression test for the None update path.
• Documents the boundary hardening in the changelog.

Confidence Score: 5/5

This looks safe to merge.

• No blocking issues found in the changed code.

Important Files Changed

Filename	Overview
skills/last30days/scripts/store.py	Normalizes incoming engagement scores before comparing them with existing stored values.
tests/test_store.py	Adds coverage for a re-sighted finding whose engagement score is None.
CHANGELOG.md	Records the store_findings null-safety fix.

_{Reviews (1): Last reviewed commit: "fix(store): guard new_engagement against..." | Re-trigger Greptile}