arch(ws-11): add loop input port

[henrypark133]

Context

Input-drain host port branch. It gives the executor a durable/caller-visible way to drain steering and follow-up inputs without giving strategies direct queue access.

Master spec: docs/reborn/agent-loop-skeleton.md
Workstream brief: docs/reborn/agent-loop-briefs/loop-input-port.md
Stack base: arch/ws-8

Latest stack maintenance on 2026-05-14:

• Rebased this branch onto its current stack base after the WS0 prompt-authority fix and the follow-up WS8/WS14-parent/WS16/WS17 conflict resolutions.
• Pushed the updated branch with force-with-lease where the remote already existed, or published it as a new branch where it did not.
• Verified the final ancestry chain from origin/reborn-integration through WS17 before publishing the PR descriptions.

What landed

• Loop input cursor and drain request/response types in the host contract.
• Host-managed input port implementation for steering/follow-up queues.
• Executor support for applying InputDrainStrategy decisions through the host port.
• Test-support adapters that capture the exact input-drain arguments passed by production callers.
• Caller-level tests around planned driver input drain behavior.

Reviewer focus

• Strategies should decide when to drain; the host port owns how inputs are loaded and marked observed.
• Input cursors should advance explicitly and survive checkpoint/resume boundaries.
• Mocks must capture every argument production code passes through the host API.

Non-goals / deferred work

• Cancellation observation is WS13.
• Progress events are WS12.
• Product inbound message cutover is WS17.

Validation

• cargo check -p ironclaw_reborn
• cargo test -p ironclaw_loop_support --lib input_port
• git diff --check
• Included in arch/ws-14-parent validation: multi-crate cargo check with ironclaw_agent_loop/test-support.

Stack position

[#3550 ws-0] state/checkpoint foundation -> reborn-integration
   |-- #3551 ws-1 strategy alpha -> ws-0
   |-- #3552 ws-2 strategy beta -> ws-0
   |-- #3553 ws-3 strategy gamma -> ws-0
   |-- #3643 ws-3.5 loop family registry -> ws-0
   '-- #3554 level1-merged -> ws-0
         |-- #3555 ws-4 planner facade -> level1
         |-- #3556 ws-5 default strategies -> level1
         '-- #3557 level2-merged -> level1
               '-- #3596 ws-6a canonical executor -> level2
                     '-- #3597 ws-7 PlannedDriver adapter -> ws-6a
                           '-- #3598 ws-8 integration/test support -> ws-7
                                 |-- #3644 ws-9 capability host wiring -> ws-8
                                 |-- #3645 ws-10 checkpoint load/resume -> ws-8
                                 |-- #3646 ws-11 input port -> ws-8
                                 |-- #3647 ws-12 progress port -> ws-8
                                 |-- #3648 ws-13 cancellation accessor -> ws-8
                                 |-- #3649 ws-15 prompt/identity context -> ws-8
                                 '-- #3650 ws-14-parent integrated host ports -> ws-8
                                       '-- #3651 ws-14 planned default registration -> ws-14-parent
                                             '-- #3652 ws-16 live runtime wiring -> ws-14
                                                   '-- #3653 ws-17 product live cutover -> ws-16

[gemini-code-assist]

Code Review

This pull request introduces a host-owned input queue system for the Reborn loop, including the HostInputQueue trait and the HostQueueLoopInputPort adapter. It also updates the RebornLoopDriverHostFactory to integrate this new input mechanism and adds comprehensive tests for steering and follow-up message handling. Feedback suggests implementing a maximum batch size limit in poll_inputs to ensure consistency and prevent performance issues from excessively large batches.

[serrrfirat]

Summary

Reviewed WS11 PR #3646 only.
Base cbdd70890afb200cdfe47076b23a885f0c70cd8c → head 4bb9af1b4c16524cf8fe3854574b4b0be2979d21.

Loop input port run-binding looks good, but ack identity is cursor-through rather than per-input token, breaking mixed-batch replay/idempotency invariant.

Validation:

• cargo check -p ironclaw_reborn ✅

Findings

#	Sev	Category	File:Line	Issue	Fix suggestion
1	High	Correctness / Replay / Idempotency	crates/ironclaw_loop_support/src/input_queue.rs:32-42, crates/ironclaw_loop_support/src/input_port.rs:53-58	ack_inputs acks through a cursor, so host can only acknowledge a contiguous prefix. Mixed batches with control input before/among user-facing inputs force unsafe choice: over-ack unseen control input, or leave already-consumed message redeliverable. WS11 brief calls for per-input ack tokens for this hazard.	Replace cursor-through ack with exact per-input ack tokens/ids. Add mixed-batch caller-boundary tests like [Cancel, UserMessage] and [GateResolved, FollowUp].

Security/data-flow notes

• Poll/ack run cursor validation is present.
• Input payloads are refs/control enums; no raw-content leak found in changed path.

Correctness/invariant notes

• Replay/idempotency boundary needs exact consumed-input identity, not cursor only.

Missing tests

• Mixed batch with early control input and later consumable input proves no over-ack/no redelivery.
• Driver/executor caller-boundary replay test, not adapter-only homogeneous batches.

[henrypark133]

Follow-up from the ws-6a review pass on #3596:

#3596 now defers user-facing input ack until after the checkpoint containing the advanced cursor is durable. This WS11 input-port PR still owns the more precise production input contract.

Please make sure this branch preserves the same ordering with the real input implementation: durable cursor/checkpoint first, physical ack second. If the real port supports mixed batches, exact per-input ack tokens should avoid acknowledging an entire batch just because one user-facing item was consumed; advance only the contiguous durable prefix that the checkpoint can recover from.

[gemini-code-assist]

Warning

You have reached your daily quota limit. Please wait up to 24 hours and I will start processing your requests again!

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 00f8dfad20

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".