Record ignore_actors refusals so ignored spam is auditable

[varunsood189]

Summary

Subscription ignore_actors dropped matching events quietly. No refusal was written and no decision appeared in the overnight report, so deliberate ignores looked like silence.

This PR records an ignore_actors refusal + decision (with reason) and returns ignored in the process outcome, without starting an agent.

What was broken

1. Actor on ignore_actors was filtered out with no ledger entry.
2. Morning report / refusals could not show why spam was ignored.
3. Other controls (self_trigger, budgets, etc.) already recorded refusals — ignore_actors did not.

Fix

• File: s16code/events/engine.py
• On ignore_actors match: record_refusal(control="ignore_actors", ...), add_decision(...), include in decisions, set ignored count.
• No agent / relevance LLM call for ignored actors.

Proof

Check	Result
Unit	uv run pytest tests/test_autonomous_events.py::test_ignore_actors_are_recorded_and_do_not_start_agents -q → passed
Behavior	ignored actor never reaches relevance gate; refusals[0].control == "ignore_actors"

Diff is only engine + test. No .env, tokens, or real channel traffic.

Test plan

• uv run pytest tests/test_autonomous_events.py::test_ignore_actors_are_recorded_and_do_not_start_agents -q
• Confirm ignored actor does not start an agent
• Confirm refusal control is ignore_actors and reason names the actor
• Confirm PR files are only:
  • s16code/events/engine.py
  • tests/test_autonomous_events.py

Made with Cursor

[theschoolofai]

Session 16 — graded ✅

Score: +100.

ignore_actors refusals were dropped rather than recorded, so the loop-protection the session teaches left no audit trail. An assistant that ignores something must be able to show it did.