Cover reviewed MIP-3, MIP-4 and MIP-7 knowledge and Markdown responses

[natureloved]

Closes #57.

Adds regression coverage for the reviewed MIP-3, MIP-4 and MIP-7 facts in the chat knowledge bundle. Strengthens the MIP-3 Markdown response checks and adds MIP-4 coverage in the route tests, including source links and absence of placeholders.

The source documents and reviewed-source metadata are already on master and remain unchanged. Assertions use the current MIP-7 Review status.

[vercel]

@natureloved is attempting to deploy a commit to the MF Experimental Team on Vercel.

A member of the Team first needs to authorize it.

[portdeveloper]

Two things to fix before this lands:

1. content/mips/upstream-sources.json keeps MIP-12 on the old Draft baseline solely to satisfy scripts/check-mip-upstream.test.ts. Its Final status and activation were already reviewed in #56. Please advance that entry to the reviewed source and give the Draft-to-Final test its own fixture manifest. Keeping production metadata stale for a test leaves the scheduled checker permanently reporting known drift. Also record the review disposition for each MIP in the PR, including any remaining UI mismatch and its follow-up, as #57 asks.

2. _overview.md calls MIP-3 an "8 MB call-frame memory limit", and the detailed heading repeats that wording. The canonical rule is a shared 8 MB active-memory pool across nested frames. Please say that consistently: a parent holding 3 MB leaves at most 5 MB for its child, rather than each frame having an independent 8 MB allowance. The detailed pool-accounting paragraph already explains this correctly.

I also approved the waiting first-contributor CI run so the checks can execute.

[natureloved]

Thanks for the review. Both items have been addressed in commit f4fcb6b:

1. MIP-12 manifest & test fixture: Advanced mip-12 in content/mips/upstream-sources.json to the reviewed commit (2a7e18894f1e55fb043080cb8cef15c7f5647768) and blob SHA (114c7e0a3c48b2e964a3d89e48e57bc9060127f8). Provided a dedicated fixture manifest in scripts/check-mip-upstream.test.ts so the offline Draft-to-Final test passes without leaving production metadata stale. npm run check:mip-upstream now reports MIP-12 as Unchanged against live upstream.
2. MIP-3 shared pool consistency: Updated _overview.md and content/mips/mip-3.md (heading and body) to consistently describe the rule as a shared 8 MB active-memory pool across nested frames, explicitly clarifying that a parent holding 3 MB leaves at most 5 MB for its child rather than each frame having an independent allowance.

Upstream Review Disposition & UI Status

MIP	Upstream Spec Diff Disposition (d2fe408...b49e903)	UI Status & Follow-Up
MIP-3	Clarified that exceeding the remaining memory limit halts exceptionally consuming all remaining gas (rather than a Solidity revert).	Interactive memory pool visualizer and cost calculator match the 8 MB pool and linear pricing schedule. No UI mismatch.
MIP-4	Formalized constants (GAS_DIPPED_INTO_RESERVE = 100, SELECTOR_DIPPED_INTO_RESERVE = 0x3a61584e), CALL-only invocation semantics (STATICCALL/DELEGATECALL/CALLCODE revert), strict evaluation order of calldata validation failures with error messages, and all-gas-consuming revert behavior.	UI Mismatch: The existing interactive bundler component does not simulate the full multi-operation Bundle Rescue flow mid-execution. This is tracked and owned by follow-up issue #35 ("Redesign the MIP-4 bundler comparison as a Bundle Rescue demo").
MIP-7	Sourced trailing EXTENSION byte rule (causes exceptional halt, consuming all remaining gas), immediate argument operand formats (restricted-range vs PUSH-prefix), undefined selector behavior (INVALID / 0xFE), and selector exclusion rules (0x5B JUMPDEST, 0x60-0x7F PUSH1-PUSH32). No network activation evidence exists in this bundle. (Upstream commit da84fa2 subsequently marked status as Review without spec alterations).	Static opcode specification table matches; no UI mismatch.
MIP-8	Status updated Draft $\rightarrow$ Final; formatted pseudocode and exponents. Sourced activation under MONAD_TEN upgrade on September 2, 2026.	Page-ified storage watch and reference data structures aligned; no UI mismatch.
MIP-12	Status updated Draft $\rightarrow$ Final. Sourced activation at mainnet round 89,758,000 (July 23, 2026) in #56. Advanced manifest to reviewed commit 2a7e188....	Parameter cards and timeline match; no UI mismatch.

[portdeveloper]

The two points from my previous review are addressed. This now needs a rebase onto current master before it can land: the source documents and metadata changed while this PR was waiting.

Please preserve the newer reviewed content when resolving the conflicts. In particular, content/mips/mip-7.md and _overview.md still say Draft on this branch, and the new assertions in src/lib/ai/knowledge.test.ts and src/lib/mip-markdown.test.ts expect Draft. Current master records Review and includes the MIP-18 context. Update those assertions to the current facts.

Keep the current entries for MIP-11, MIP-15 and MIP-18 in content/mips/upstream-sources.json, along with the reviewed source revision already on master. The overlapping MIP-3 and MIP-4 documents have also been filled in there. Please narrow the remaining diff to useful additions and regression coverage, then run the knowledge-bundle, Markdown-route and upstream-checker tests on the rebased result.

[natureloved]

Rebased onto current master (05d347f).

Kept master's newer content throughout: the source documents, content/mips/upstream-sources.json (including the MIP-11/MIP-15/MIP-18 entries and the reviewed 282d1812 revision), and the MIP-10/11/15/16/18 regression tests are unchanged.

The remaining diff is narrowed to the missing regression coverage:

• src/lib/ai/knowledge.test.ts: new MIP-3, MIP-4 and MIP-7 knowledge-bundle tests asserting the reviewed facts and absence of placeholders.
• src/lib/mip-markdown.test.ts: added the MIP-4 .md route test and strengthened the MIP-3 route test.
• Assertions updated to current facts (MIP-7 Review, pinned revision 282d1812).

Both points from the previous review remain addressed through master's newer content: the MIP-12 manifest entry is already advanced, and MIP-3 is consistently described as a shared 8 MB active-memory pool across nested frames.

Verification on the rebased result:

• knowledge-bundle, Markdown-route and upstream-checker tests: 59/59 pass (full suite 160/160).
• pnpm check:mip-upstream against live upstream: all entries Unchanged.

[portdeveloper]

lgtm, thanks for rebasing and keeping the current source content intact