feat(autogen): add AG2-018, tool prints to stdout for diagnostics

[bradAGI]

Ports OAI-010 to AutoGen — the observability dimension isn't covered in any of the five newer packs.

AutoGen makes this especially deceptive, which is what the rule text leads with. The conversation itself is narrated to stdout, so a tool print lands in the middle of the transcript and reads as though it were part of the exchange — when in fact it's invisible to the agents reasoning over it and carries no marker for which round or which tool call produced it. Anyone reconstructing the run from a log sink rather than a terminal loses it entirely.

The fix points at the return value as "the only channel the agents actually read," which is the distinction that matters in a two-agent reply loop.

Verification — engine built at main:

$ trustabl rules validate .
OK: 86 rule pack(s), 207 rule(s) valid under rule schema version 14

Fire (print(f"looking up {order_id}") in a tool registered via register_for_llm/register_for_execution): AG2-018, AG2-201
Silent (logger.info("looking up %s", order_id)): AG2-201

has_print_call matches a bare print callee, so pprint and other attribute calls don't false-positive.

No new predicates, so no schema_version bump.

[jhumel-code]

Thanks @bradAGI, this is a model contribution. I checked the rule against the engine at main: has_print_call exists in schema.go, predicates.go, and the evaluator, autogen_tool is a valid tool-scope applies_to, and you are right that no schema_version bump is needed since no new predicate is introduced. Severity and confidence sit exactly where the neighbors do (OAI-010 is low/0.65, ADK-009 low/0.7), and the pack id autogen_observability follows the <category>_<topic> convention. I also appreciate that the explanation is genuinely AutoGen framing rather than a copy of the OAI-010 text; the point about a print landing mid-transcript and reading as though it were part of the exchange is the right consequence to lead with. Picking AG2-018 to steer clear of the open PRs contesting AG2-013 through AG2-017 was thoughtful, and the ID is unclaimed by any other open PR. The engine half at trustabl/agent-reliability-analyzer#153 mirrors this file byte for byte, so rules-sync stays green as long as the two land together. One housekeeping note: every shipped rule needs a rationale doc in trustabl-rulebook before release; we can handle that side or pair on it, it is not on you to write. Approving, and I will coordinate the paired merge.