Thanks for the context here. I did a careful shell check against current main, and this is already implemented.

Close: current main now has same-or-better bounded JSON response handling across the touched extension paths, and the proof commit is current-main-only rather than shipped in v2026.6.11.

So I’m closing this as already implemented rather than keeping a duplicate issue open.

Current CI triage for this head (66e9df36e13d8194030f8d2b8c505bc3ed49a231): check-lint is red, and one failure is PR-local.

• extensions/feishu/src/streaming-card.test.ts:68:13 reports no-unused-vars for ok. That looks introduced by the test mock conversion to real Response objects: Response.ok is now derived from status, so the local ok variable and ok = false assignments can be removed while keeping the existing status handling.
• The same lint job also reports extensions/line/src/message-cards.test.ts:333:9 (loneHighSurrogate shadowing), but that file is not in this PR's changed-file list, so I would treat it as unrelated current-main lint noise unless a rebase changes the diff.

Other relevant early checks I saw were green, including Real behavior proof, check-prod-types, check-test-types, check-additional-extension-channels, and check-dependencies; several matrix jobs were still running when I checked.

Addressed both findings:

1. Feishu guard cleanup - Both token and card-create paths now wrap readProviderJsonResponse in try/finally so release() is always called, even on oversized or malformed responses.

2. Overflow proof - Added a focused test that sends a 20MB response and verifies:

   • The bounded reader rejects with a cap-exceeded error
   • The SSRF guard release() is still called via try/finally

Addressed all findings:

1. PluralKit fixture — Updated success test to return a real Response object.
2. Feishu guard cleanup — Both paths use try/finally for release().
3. Real behavior proof — Terminal output below shows the bounded reader rejecting a 20MB response with "JSON response exceeds 16777216 bytes" while normal responses parse correctly.

@clawsweeper re-review

🦞🧹
ClawSweeper re-review requested.

I asked ClawSweeper to review this item again.
Action: item re-review queued (workflow sweep.yml, event repository_dispatch).
Result: the existing ClawSweeper review comment will be edited in place when the review finishes.

Addressed remaining findings:

1. MS Teams sibling reads — Bounded response.json() in attachments/graph.ts (2 call sites).
2. Tlon scry reader — Bounded response.json() in urbit/channel-ops.ts.
3. PluralKit fixture — Updated to real Response object.

Changed-path proof — Terminal output shows readProviderJsonResponse with the "Discord PluralKit response" label (the exact label used in the changed plugin path):

• 20MB response: REJECTED: Discord PluralKit response: JSON response exceeds 16777216 bytes
• Normal response: PARSED: {"id":"456","member":{"id":"m2","name":"Alex"}}

@clawsweeper re-review

🦞🧹
ClawSweeper re-review requested.

I asked ClawSweeper to review this item again.
Action: item re-review queued (workflow sweep.yml, event repository_dispatch).
Result: the existing ClawSweeper review comment will be edited in place when the review finishes.

Narrowed PR scope: removed MS Teams to keep the PR clean and focused. MS Teams has additional sibling paths (Bot Framework attachment reader, Graph error visibility) that deserve their own dedicated PR.

This PR now covers: Feishu, Discord, Google Meet, and Tlon only.

@clawsweeper re-review

🦞🧹
ClawSweeper re-review requested.

I asked ClawSweeper to review this item again.
Action: item re-review queued (workflow sweep.yml, event repository_dispatch).
Result: the existing ClawSweeper review comment will be edited in place when the review finishes.

The focused Tlon scry slice has landed via #100376 after bounded loopback proof, sanitized AWS focused tests, exact-head CI, and fresh autoreview.

Please remove these overlapping files from this broader PR before further review:

• extensions/tlon/src/urbit/channel-ops.ts
• extensions/tlon/src/urbit/channel-ops.test.ts

The Discord, Feishu, Google Meet, Tlon API, and SDK-surface work remains distinct and should stay independently reviewable. Thanks for the broader response-bounds work.

ClawSweeper applied the proposed close for this PR.

• Action: closed this PR.
• Close reason: already implemented on main.
• Evidence: durable ClawSweeper review.