fix(j5): agent-to-human sends must be asks or replies

[Jacksondr5]

Posted by an AI agent on Jackson's behalf

Plain human-directed sends created durable ghost traffic: no human surface and no accountable obligation. Product has ruled that behavior out permanently.

This PR refuses a human-targeted send unless it is an ask (expect_reply=true with intent and urgency) or a reply (exchange_id). It also refuses explicit human exchange follow-ups before any ledger write. A second ask to a person while one is already open is refused with a typed error instead of silently coalescing. Agent-to-agent plain sends remain open. Human participant rows now expose can_receive_message=false and can_open_exchange=true, and unreachable-participant recovery text now distinguishes an addressable agent from a human row that can open an ask. The served version-15 tool description is mechanically byte-equal to the product contract.

Tests: focused A2A service, inbox, formatter, toolkit, handler, and registration suites (48 passed); vp run --filter t3 typecheck (no errors).

Built with GPT-5.6 Codex in T3 Code.

Closes #44

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on this repository. Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Essentials

Run ID: 0d0554df-db08-43bd-ab89-82b0afa497ed

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

---

_{Comment @coderabbitai help to get the list of available commands.}

[Jacksondr5]

Posted by an AI agent on Jackson's behalf

CodeRabbit review disposition at c630dc26:

• Major — confirmed and fixed in 2fa7749ae. A human-directed exchange follow-up was neither an ask nor a reply. The send path now rejects both explicit exchange_id follow-ups and implicit expect_reply=true follow-ups after an open human ask, before any ledger write; both shapes have no-write regression coverage.
• Minor — confirmed and fixed in 2fa7749ae. The unreachable-participant recovery text now distinguishes an addressable agent (canReceiveMessage=true) from a human row that can open an ask (canOpenExchange=true).

Local verification at 2fa7749ae: six focused A2A/MCP suites, 34 tests passed; vp run --filter t3 typecheck passed (pre-existing suggestions only). Fresh exact-head CI is pending.

[Jacksondr5]

Posted by an AI agent on Jackson's behalf

Post-review CI disposition: both exact-head workflows at 2fa7749ae correctly exposed dependent test fixtures, not a production regression. 138b8513b updates the exact recovery-response assertion and represents the delivery race as a historical persisted follow-up, so the worker preserves compatibility for already-durable traffic while send_message refuses new human follow-ups.

Local verification at 138b8513b: 8 focused A2A/MCP suites, 48 tests passed; vp run --filter t3 typecheck passed (pre-existing suggestions only). Fresh exact-head CI is pending.