docs: full re-audit of the agent doc graph and knowledge base

[Roasbeef]

Overview

The client side of the same doc sweep. A lot has changed recently, the
per-package agent docs and the docs/ knowledge base had drifted, and
make doc-check was failing (diverged CLAUDE.md/AGENTS.md pairs and two
docs missing from the index). This PR does one full-repo sweep: re-audit every
doc against current source, document the packages that had none, and get the
whole graph accurate and green.

This supersedes the stale nightly sweep in #898, which can be closed.

Docs-only. No .go files change.

What's in here

Four commits, each a self-contained concern:

1. Root guide fixes. The highest-traffic doc referenced a nonexistent
   make itest target (integration tests run under make systest), pointed at
   db/queries//db/schema/ instead of db/sqlc/queries//db/sqlc/migrations/,
   and cited a .editorconfig that does not exist in this repo.
2. Sixteen new package docs. Packages that had no agent doc now have one
   (credit, coinselect, chainfees, rpcauth,
   chainbackends/lndsubmitter, internal/sqlbase, sdk/walletdk/mobile,
   cmd/walletdk-wasm, and the generated proto/sqlc stubs). Substantive
   packages are slotted into the ARCHITECTURE.md layer tables; generated
   stubs stay under the existing rpc/mailbox rollup rows.
3. Per-package re-audit. Every existing CLAUDE.md/AGENTS.md was checked
   against source: drifted types, message directions, and dependency edges
   corrected, and a leaked tool-output artifact from a prior automated sweep
   stripped out. This pass also restored a set of load-bearing oor invariants
   (the session-actor deadlock rule, dedup/commit-race semantics, and the
   lineage/ancestry bounds) that an over-eager tightening had cut and that
   lived nowhere else.
4. Knowledge-base tighten. Concrete claims across docs/ verified against
   source and fixed, the two orphaned design docs added to the index, and
   cross-repo relative links that cannot resolve neutralized to plain text.

How it was done

The re-audit ran as a parallel Sonnet fan-out (one worker per small batch of
packages), each grounding its edits in go doc and actual source. Shared files
(ARCHITECTURE.md, docs/index.md, root CLAUDE.md) were reconciled
centrally so parallel workers never collided, and make doc-check gates the
four cross-link invariants.

A final independent review pass (a stronger model reading the whole diff)
verified claims against source and flagged eight corrections, every one folded
in after re-confirming it against the code (for example: internal/sqlbase is
a single <prefix>_kv table with parent_id nesting, not a table-per-bucket
scheme; SendOnChainStatus is Submitted/Preview, not DryRun; and the
chainbackends PackageSubmitter LND fallback to lndsubmitter.New).

make doc-check passes.

Known follow-ups (out of scope for a docs-only PR)

• A couple of stale .go doc-comments the sweep noticed but did not touch
  (sdk/walletdk/mobile/doc.go references a stub.go that does not exist;
  arkrpc/treeconv is currently dead code).
• Three design docs describe work that has since shipped and want a dedicated
  status rewrite: canonical_activity_log_design.md (issue swapwallet: canonical append-only activity record with stable cross-restart ids #774 largely
  landed), swap_system.md (a third forfeit RPC and the CREDIT/MIXED
  settlement types), and fee-change-model.md.

[gemini-code-assist]

Code Review

This pull request performs a comprehensive documentation update across the entire repository, synchronizing AGENTS.md, CLAUDE.md, and other design documents in the docs/ directory with recent architectural changes. Key updates reflect the integration of the client-side credit subsystem, the mobile SDK facade, the WASM build target, protocol version negotiation, and various refactorings such as the transition from lib/scripts to lib/arkscript. I have no further feedback to provide as there are no review comments.

Important

The consumer version of Gemini Code Assist on GitHub is being sunset. Starting June 18, 2026, new organization installations will be blocked, and all code review activity will officially cease on July 17, 2026.
For more details on the timeline and next steps, please review the Help Documentation.