bug: Invalid tags silently dropped in Belief model validator, breaks future filtering (post-Dec 5 commit)

[moonrunnerkc]

Context

After pulling latest (Dec 5 model refinements), the Belief Pydantic validator for tags (List[str]) silently filters out non-string items during creation, which broke a quick test of tag-based listing in storage stubs. This propagates to agent tagging (e.g., PerceptionAgent) without warnings.

Steps to Reproduce

1. git pull origin main (post-Dec 5).
2. In a test script: belief = Belief(..., tags=["user_pref", None, 123]) → Create via store.
3. await store.list(tags=["user_pref"]) stub → Misses belief (tags became ["user_pref"] only).
4. Check logs → No error; just quiet drop.

Expected vs Actual

• Expected: Validator raises ValidationError on invalid tags (e.g., None/int) for explicit debugging.
• Actual: Strips invalids silently → Subtle data loss, hard to trace in agent pipelines.

Logs/Screenshots

• Pydantic log (if verbose): WARNING: Ignoring invalid tag 'None' in list.
• Test output: showing len(tags) mismatch.
• Commit ref: Likely in belief.py @root_validator for tags.

Suggested Fix

Update validator to raise ValueError on non-str tags. Add a unit test in tests/core/models/test_belief.py for tag validation. Fix before Phase 4 agents rely on tags.