docs: the stdio protocol doc describes the wire that ships

[Max17190]

Why

docs/stdio-protocol.md calls itself the normative reference for every field of every line and documents openmax-stdio/3. The binary speaks /5, and openmax --check --stdio rejects a /3 hello on its first line, so a client written from the doc fails conformance before it sends a command. Two events and two stop reasons that core emits today are named nowhere a client author would look.

Summary

• docs/stdio-protocol.md now matches openmax --spec stdio: the hello line carries continued; the event table gains harness_note, compacted, turn_refused, and the optional approval_request.env with the rendering rule for it; the stop_reason table gains budget_exhausted and unverified; the process exit code is stated; one "what changed" section per bump, in the order the bumps happened.
• The printed contract (spec.rs) omitted continued from its hello line while the binary emits it.
• The Done doc comment on the wire type listed six stop reasons; the loop emits eight. All eight are described.
• The stdio module doc pointed at a README section that does not exist; it now points at the doc and the spec.
• README named the /3 contract twice.
• The usage doc referred to a delegate skill the tree stopped shipping in chore(skills): the tree's extension surfaces ship empty #240.
• The configuration example omitted context_tokens, which the text two paragraphs later calls required.

No behavior change: documentation, one string in the printed spec, and two doc comments.

Test Plan

• cargo clippy --workspace --all-targets --locked -- -D warnings: clean.
• cargo test --workspace --locked: 944 passed, 11 ignored, 0 failed. The stdio spec drift tests and the wire golden pass unchanged.
• Every field, stop reason, and exit code in the doc was checked against crates/tui/src/stdio.rs, crates/tui/src/headless.rs, crates/core/src/agent.rs, and the output of openmax --spec stdio from a release build.

Greptile Summary

Updates the stdio protocol reference and related documentation for the shipped openmax-stdio/5 contract. The documented /5 handshake, legacy /3 rejection, harness_note and turn_refused envelopes, and provider-error completion behavior were exercised against the CLI and matched the documentation. No defects were found.

Confidence Score: 5/5

Safe to merge based on the exercised stdio protocol and error-handling flows.

The final review contains no findings: the live CLI behavior and generated protocol validation agreed with the updated documentation.

Files Needing Attention: None. The changed documentation files were consistent with the tested CLI contract.

T-Rex Logs

What T-Rex did

• The saved validation script was executed successfully.
• The logs show the shipped CLI rejects an openmax-stdio/3 hello as unsupported, then emits openmax-stdio/5 and protocol version 5 during a live stdio session.
• The validation confirms that the newly documented harness_note and turn_refused envelopes are accepted through --check --stdio.
• When a provider connection failure occurs, the CLI exits with code 1 after emitting error and done, with stop_reason: error.
• The observed behaviors align with the updated documentation.

_{Ran code and verified through T-Rex}

_{Reviews (1): Last reviewed commit: "docs: the stdio protocol doc describes t..." | Re-trigger Greptile}