Add token2022-integration-guard skill (consumer-side Token-2022 safety)

[mohit-1710]

Adds token2022-integration-guard as .claude/skills/ext/token2022-integration-guard, with one routing line in the hub under ## Token Extensions. It is the consumer/integrator counterpart to the existing token-2022.md (issuer-side).

The kit teaches how to issue and inspect Token-2022 tokens. This is what a program actually has to do when it must accept an arbitrary, possibly hostile mint (AMM, vault, lending, escrow) without getting drained or bricked. Inspectors tell you a mint is hostile. This guard refuses it on-chain.

What it ships:

• a compiling Anchor 1.0 guard vault: an on-chain extension allowlist that rejects hostile mints (permanent delegate, transfer hook, frozen-default, confidential), a delta-measured fee-safe deposit, and a reentrancy-aware withdraw
• a t22-scan CLI with a per-profile verdict and a real CI exit code, verified against live mainnet (PYUSD rejects on an active permanent delegate, USDC allows)
• 6 LiteSVM attack tests that load the real spl_token_2022 program and prove each guard blocks its attack

Reproduce:

cd .claude/skills/ext/token2022-integration-guard/examples/guard && anchor build --ignore-keys && bun run test   # 6 passing

Changes: the submodule + .gitmodules entry, one routing line in .claude/skills/SKILL.md, and the ext/ count in the README (18 to 19). Follows the kit shape (skill/SKILL.md router + focused files + agents/commands/rules + install.sh). MIT.

Repo: https://github.com/mohit-1710/token2022-integration-guard

[vercel]

@mohit-1710 is attempting to deploy a commit to the STBR TRUE Team on Vercel.

A member of the Team first needs to authorize it.