docs(coding-agent): advisor, prompt-settlement, and ask-user architecture docs

[DankerMu]

Summary

Adds three architecture design docs (status: direction accepted, pre-implementation), split along delivery boundaries so each can be reviewed and shipped independently:

• advisor-architecture.md — root-runtime-owned, read-only reviewer subruntime (read/grep/glob/git_read over a new thin FileSystemView), candidate-terminal-triggered reviews, severity routing (nit/concern/blocker), bounded correction loops with a prompt-admission sealing rule, fail-closed enable / fail-open runtime faults.
• prompt-settlement.md — session-level PromptSettlementTracker + structured PromptOutcome across TUI/Print/JSON/RPC/ACP; agent_end keeps its run-level meaning, settlement becomes a separate contract (also fixes the semantics gap where RPC promptAndWait resolves on the first agent_end).
• ask-user.md — universal ask_user session capability: live responders in-run, persistent fallback pending queues with needs_user_input outcomes, and a shared TUI editor-surface FIFO dialog arbiter (fixes an existing dialog-clobbering hazard).

All source-code contract claims in the docs were verified against the current tree (section 20 of each doc). Delivery is planned as six independently verifiable slices (arbiter → settlement → ask_user ∥ advisor core → TUI/daemon surface → persistence/ops).

Also registers the new docs (plus the existing continual-harness roadmap page) in docs.json and index.md navigation.

Notes

• Docs-only change; no runtime code touched.

🤖 Generated with Claude Code

Note

Add architecture docs for Advisor, Prompt Settlement, and ask_user to coding-agent

Adds three architecture design documents to the coding-agent docs:

• advisor-architecture.md covers the Advisor subsystem lifecycle, severity taxonomy, review triggering, mode behaviors (TUI/daemon/Print/JSON/RPC/ACP), SessionAdvisorManager, and phased rollout plan.
• prompt-settlement.md defines session-level prompt settlement tracking, stable promptId lineage across retries, the PromptOutcome contract, and per-mode behavior including exit code 2 for needs_user_input in Print mode.
• ask-user.md documents the ask_user capability contract, UserInputBroker architecture, TUI pending queue/FIFO dialog arbiter, daemon multi-responder semantics, persistence/recovery, and phased rollout.
• Both docs.json and index.md are updated to register the new pages in the docs site navigation.

^{Macroscope summarized 698c231.}

[github-actions]

Hi @DankerMu, thanks for your interest in contributing!

This project requires that pull request authors are vouched, and you are not in the list of vouched users.

This PR will be closed automatically. See https://github.com/PrimeIntellect-ai/prime-agent/blob/main/CONTRIBUTING.md for more details.