fix(google): bound OAuth project and token JSON response reads by hugenshen · Pull Request #97587 · openclaw/openclaw

hugenshen · 2026-06-28T23:14:14Z

What Problem This Solves

extensions/google/oauth.project.ts and extensions/google/oauth.token.ts contained six
unbounded await response.json() calls on the Gemini CLI OAuth onboarding and token-exchange
paths (getUserEmail, pollOperation, loadCodeAssist success + error, onboardUser,
requestTokenGrant).

After fetchWithTimeout materialises the response body, response.json() can trigger a
second large allocation during JSON.parse. Replacing these calls with
readProviderJsonResponse caps the JSON parse step at 16 MiB with labeled errors at each
OAuth call site.

This complements the shared fetch boundary landed in #97628
(readResponseWithLimit inside fetchWithTimeout). Together the two layers give defense in
depth:

Fetch boundary (fix(google): bound google OAuth fetchWithTimeout arrayBuffer at 16 MiB #97628 on main): bounds upstream stream reads before buffering.
Parse boundary (this PR): bounds JSON.parse on the buffered Response objects that
OAuth call sites consume.

Changes

extensions/google/oauth.project.ts:
- imports readProviderJsonResponse from openclaw/plugin-sdk/provider-http
- replaces six response.json() reads with labeled readProviderJsonResponse(...) calls
  (google.userinfo, google.poll-operation, google.load-code-assist,
  google.onboard-user)
extensions/google/oauth.token.ts:
- imports readProviderJsonResponse from openclaw/plugin-sdk/provider-http
- replaces await response.json() → readProviderJsonResponse(response, "google.token")
  in requestTokenGrant
extensions/google/oauth.test.ts:
- adds four regression tests:
  - rejects an oversized token exchange response body (end-to-end via
    exchangeCodeForTokens; fetch cap fires first after fix(google): bound google OAuth fetchWithTimeout arrayBuffer at 16 MiB #97628)
  - rejects an oversized token body at the JSON parse boundary (defense-in-depth when
    fetchWithTimeout already returned a buffered oversized body)
  - rejects an oversized loadCodeAssist success response body (via
    resolveGoogleOAuthIdentity)
  - swallows bound error on oversized userinfo body and returns undefined email

Rebased onto current main; the stale plugin SDK surface budget commit was dropped because
#97628 and other main commits already carry those values.

Real behavior proof

Behavior addressed: Oversized or hostile Google OAuth JSON responses fail closed instead
of triggering unbounded JSON.parse allocation on buffered bodies returned by
fetchWithTimeout.

Real environment tested: Node v22, macOS 24.3.0, node:http streaming server on
127.0.0.1, built helpers from dist/plugin-sdk/provider-http.js and
dist/plugin-sdk/response-limit-runtime.js.

Exact steps:

Stream an ~18 MiB OAuth token JSON body from a local HTTP server (no Content-Length).
Drive readResponseWithLimit at the fetch boundary → assert labeled overflow error.
Drive readProviderJsonResponse on the same streamed body → assert labeled parse cap.
Negative control: unbounded response.body reader consumes the full ~18 MiB stream.
Verify post-fetchWithTimeout buffered Response shape rejects at parse boundary.
Verify normal token / userinfo / loadCodeAssist bodies still parse.
Verify VPC-SC error-path .catch(() => null) swallows oversized error bodies.
Run OAuth module regression tests through exchangeCodeForTokens and
resolveGoogleOAuthIdentity.

Observed result:

Fetch boundary: google HTTP fetch: body exceeds 16777216 bytes (got 16817603)
Parse boundary (stream): google.token: JSON response exceeds 16777216 bytes
Negative control: unbounded reader consumed 18 874 405 bytes (~18 MiB)
Buffered oversized body (16 778 297 bytes) rejected before JSON.parse
Normal token / userinfo / loadCodeAssist responses parse correctly
VPC-SC error-path .catch(() => null) returns null on oversized body
ALL PROOF ASSERTIONS PASSED

What was not tested: live Google OAuth API calls.

Evidence

=== Google OAuth response bound proof (path-level) ===
Layers: (1) fetchWithTimeout readResponseWithLimit [#97628 on main]
        (2) readProviderJsonResponse on OAuth call sites [#97587]

--- Case 1: oversized streamed token body (fetch boundary) ---
  ok: readResponseWithLimit rejects oversized streamed OAuth body
  ok: fetch-boundary error is labeled (got: google HTTP fetch: body exceeds 16777216 bytes (got 16817603))
  ok: negative control: unbounded reader consumed full ~18 MiB stream (18874405 bytes)

--- Case 2: oversized streamed body rejected at JSON parse boundary ---
  ok: readProviderJsonResponse rejects oversized streamed body
  ok: parse-boundary label present (got: google.token: JSON response exceeds 16777216 bytes)

--- Case 3: oversized buffered OAuth token body (post-fetch shape) ---
  ok: buffered oversized body rejected before JSON.parse
  ok: body size 16778297 bytes exceeds 16777216 cap

--- Case 4: normal OAuth JSON responses still parse ---
  ok: token body parses
  ok: userinfo body parses
  ok: loadCodeAssist body parses

--- Case 5: loadCodeAssist error path .catch(() => null) ---
  ok: oversized VPC-SC probe body returns null via .catch

--- Case 6: OAuth module path-level regression tests ---
  Tests  4 passed | 27 skipped (31)
  ok: vitest OAuth bound regression tests passed

ALL PROOF ASSERTIONS PASSED

OPENCLAW_VITEST_MAX_WORKERS=1 node scripts/run-vitest.mjs extensions/google/oauth.test.ts \
  -t "rejects an oversized|swallows bound"
# Tests  4 passed | 27 skipped (31)

New bound regression tests (4 added):

loginGeminiCliOAuth > rejects an oversized token exchange response body ✓
loginGeminiCliOAuth > rejects an oversized token body at the JSON parse boundary ✓
loginGeminiCliOAuth > rejects an oversized loadCodeAssist success response body ✓
loginGeminiCliOAuth > swallows bound error on oversized userinfo body and returns undefined email ✓

clawsweeper · 2026-06-28T23:17:23Z

Codex review: needs maintainer review before merge. Reviewed July 4, 2026, 10:13 PM ET / 02:13 UTC.

Summary
Replaces Google Gemini CLI OAuth project/token JSON response parsing with readProviderJsonResponse and adds oversized-response regression coverage.

PR surface: Source +7, Tests +134. Total +141 across 3 files.

Reproducibility: yes. Source inspection shows raw response.json() parser sites on current main, and the PR body/diff includes oversized Response paths plus focused regression tests; I did not execute tests because this review is read-only.

Review metrics: 1 noteworthy metric.

Google OAuth JSON parser call sites: 6 replaced with bounded helper. Each changed reader sits on token, identity, project discovery, or onboarding auth-provider paths, so maintainers should explicitly accept the parser-boundary behavior change.

Root-cause cluster
Relationship: canonical
Canonical: #97587
Summary: This PR is the viable current landing path for Gemini CLI Google OAuth parser-boundary hardening; related items cover adjacent transport/error caps or broader overlapping work.

Members:

adjacent_distinct: fix(google): bound google OAuth fetchWithTimeout arrayBuffer at 16 MiB #97628 - Merged companion fix for the shared Google OAuth fetch/body buffering boundary, not the remaining parser call sites.
adjacent_distinct: fix(google): bound OAuth token error response reads #99605 - Merged token error-body cap in the same helper, separate from this PR's success/project JSON parser boundary.
partial_overlap: fix(openrouter,google): bound OAuth response body reads #97745 - Open conflicting PR overlaps Google token success parsing and also touches OpenRouter, so it is not a safe canonical replacement.
partial_overlap: fix(google): bound OAuth token error body reads to prevent OOM #98722 - Open conflicting PR overlaps Google token response hardening, but its primary error-body target is already covered by the merged token error-cap PR.
partial_overlap: fix(google): bound OAuth response reads to prevent OOM #99812 - Newer broader PR overlaps these parser replacements while also adding Vertex ADC gzip/read changes, so it should not supersede this narrower reviewed branch yet.

Proposal only: this assessment does not dispatch repair, suppress jobs, mutate sibling items, close, or merge anything.

Merge readiness
Overall: 🐚 platinum hermit
Proof: 🦞 diamond lobster
Patch quality: 🐚 platinum hermit
Result: ready for maintainer review.

Overall follows the weaker of proof and patch quality, so missing proof can cap an otherwise strong patch.

Rank-up moves:

none.

Risk before merge

[P1] Merging intentionally changes Google OAuth token/project JSON failure behavior: oversized provider JSON now fails through labeled bounded-helper errors, while userinfo continues to swallow that failure and return no email.
[P1] Several overlapping Google OAuth hardening PRs remain open, so maintainers should choose whether this narrow parser-boundary branch or a broader bundle is the canonical landing path.

Maintainer options:

Accept bounded parser failures (recommended)
Maintainers can accept the new labeled fail-closed JSON parser behavior as intended Google OAuth hardening and land this branch after normal required checks.
Pause for a broader Google OAuth bundle
If maintainers want Gemini CLI OAuth and Vertex ADC hardening reviewed together, pause this PR and choose a consolidated branch with equivalent parser coverage and proof.

Next step before merge

No automated repair is needed; the remaining action is maintainer acceptance of the auth-provider parser failure behavior and canonical choice among overlapping PRs.

Security
Cleared: No concrete security or supply-chain regression is visible; the diff uses existing SDK bounded-read helpers and does not change dependencies, workflows, permissions, or secret handling.

Review details

Best possible solution:

Land this narrow parser-boundary hardening once maintainers accept the auth-provider fail-closed behavior, and review broader Vertex ADC hardening separately unless they choose a consolidated Google OAuth bundle.

Do we have a high-confidence way to reproduce the issue?

Yes. Source inspection shows raw response.json() parser sites on current main, and the PR body/diff includes oversized Response paths plus focused regression tests; I did not execute tests because this review is read-only.

Is this the best way to solve the issue?

Yes. With the shared fetch cap and token-error cap already on main, replacing the remaining parser sites with the existing SDK bounded JSON helper is the narrowest maintainable owner-boundary fix.

AGENTS.md: found and applied where relevant.

Codex review notes: model internal, reasoning high; reviewed against 7d3cfa85dca1.

Label changes

Label justifications:

P2: This is a focused Google OAuth hardening fix with limited blast radius and no evidence of an active user outage.
merge-risk: 🚨 auth-provider: The diff changes Google OAuth token/project JSON parsing and oversized or malformed response failure behavior on auth and onboarding paths.
rating: 🐚 platinum hermit: Overall readiness is 🐚 platinum hermit; proof is 🦞 diamond lobster and patch quality is 🐚 platinum hermit.
status: 👀 ready for maintainer look: ClawSweeper has no concrete contributor-facing blocker left for this PR. Sufficient (terminal): The PR body includes after-fix terminal proof with a local streaming HTTP server, bounded helper checks, a negative control, normal-response controls, and focused OAuth regression-test output.
proof: sufficient: Contributor real behavior proof is sufficient. The PR body includes after-fix terminal proof with a local streaming HTTP server, bounded helper checks, a negative control, normal-response controls, and focused OAuth regression-test output.

Evidence reviewed

PR surface:

Source +7, Tests +134. Total +141 across 3 files.

View PR surface stats

Area	Files	Added	Removed	Net
Source	2	17	10	+7
Tests	1	134	0	+134
Docs	0	0	0	0
Config	0	0	0	0
Generated	0	0	0	0
Other	0	0	0	0
Total	3	151	10	+141

What I checked:

Root policy read: Root AGENTS.md and scoped extensions/AGENTS.md were read fully; their auth-provider review and plugin SDK import rules apply to this PR. (AGENTS.md:34, 7d3cfa85dca1)
Current main project parser sites: Current main still uses raw response.json() in Google OAuth userinfo, pollOperation, loadCodeAssist, and onboardUser parsing, so the central parser-boundary work is not already implemented on main. (extensions/google/oauth.project.ts:24, 7d3cfa85dca1)
Current main token parser site: Current main bounds token error text but still parses successful token grants with raw response.json(). (extensions/google/oauth.token.ts:39, 7d3cfa85dca1)
PR diff coverage: The PR diff imports readProviderJsonResponse and replaces the Google OAuth project/token JSON parser sites, including the token success path labeled google.token. (extensions/google/oauth.token.ts:39, 63570bb635ed)
Bounded JSON helper contract: readProviderJsonResponse reads through readResponseWithLimit, defaults to the 16 MiB provider JSON cap, labels overflow errors, and wraps malformed JSON with the caller label. (src/agents/provider-http-errors.ts:315, 7d3cfa85dca1)
Reader overflow behavior: readResponseWithLimit rejects when the prefix is truncated and the lower-level reader cancels on overflow, so the helper does not continue buffering the whole response. (src/infra/http-body.ts:259, 7d3cfa85dca1)

Likely related people:

steipete: The current split Google OAuth project/token/http module structure appears to come from refactor(google): split oauth flow modules. (role: original OAuth module contributor; confidence: high; commits: 92e765cdee15; files: extensions/google/oauth.project.ts, extensions/google/oauth.token.ts, extensions/google/oauth.http.ts)
wangmiao0668000666: Authored the merged shared Google OAuth fetchWithTimeout response-body cap that this parser-boundary PR complements. (role: recent adjacent fetch-boundary contributor; confidence: high; commits: a6aaba76ac66; files: extensions/google/oauth.http.ts, extensions/google/oauth.http.test.ts)
mushuiyu886: Authored the recently merged bounded token error-body change in the same Google OAuth token helper and tests. (role: recent adjacent token-error contributor; confidence: medium; commits: 6207a4e75bae; files: extensions/google/oauth.token.ts, extensions/google/oauth.test.ts)
vincentkoc: Prior merged work added Gemini CLI personal OAuth support in the same project/token OAuth surface. (role: adjacent OAuth behavior contributor; confidence: medium; commits: f02e43518826; files: extensions/google/oauth.project.ts, extensions/google/oauth.token.ts, extensions/google/oauth.test.ts)

What the crustacean ranks mean

🦀 challenger crab: rare, exceptional readiness with strong proof, clean implementation, and convincing validation.
🦞 diamond lobster: very strong readiness with only minor maintainer review expected.
🐚 platinum hermit: good normal PR, likely mergeable with ordinary maintainer review.
🦐 gold shrimp: useful signal, but proof or patch confidence is still limited.
🦪 silver shellfish: thin signal; proof, validation, or implementation needs work.
🧂 unranked krab: not merge-ready because proof is missing/unusable or there are serious correctness or safety concerns.
🌊 off-meta tidepool: rating does not apply to this item.

Shiny media proof means a screenshot, video, or linked artifact directly shows the changed behavior. Runtime, network, CSP, and security claims still need visible diagnostics.

How this review workflow works

ClawSweeper keeps one durable marker-backed review comment per issue or PR.
Re-runs edit this comment so the latest verdict, findings, and automation markers stay together instead of adding duplicate bot comments.
A fresh review can be triggered by eligible @clawsweeper re-review comments, exact-item GitHub events, scheduled/background review runs, or manual workflow dispatch.
PR/issue authors and users with repository write access can comment @clawsweeper re-review or @clawsweeper re-run on an open PR or issue to request a fresh review only.
Maintainers can also comment @clawsweeper review to request a fresh review only.
Fresh-review commands do not start repair, autofix, rebase, CI repair, or automerge.
Maintainer-only repair and merge flows require explicit commands such as @clawsweeper autofix, @clawsweeper automerge, @clawsweeper fix ci, or @clawsweeper address review.
Maintainers can comment @clawsweeper explain to ask for more context, or @clawsweeper stop to stop active automation.

Review history (4 earlier review cycles)

reviewed 2026-07-02T15:08:04.372Z sha 09fef5a :: needs maintainer review before merge. :: none
reviewed 2026-07-04T14:10:37.598Z sha d7da56d :: needs maintainer review before merge. :: none
reviewed 2026-07-04T16:27:43.013Z sha 63570bb :: needs maintainer review before merge. :: none
reviewed 2026-07-04T16:33:39.025Z sha 63570bb :: needs maintainer review before merge. :: none

hugenshen · 2026-07-02T14:55:09Z

@clawsweeper re-review

clawsweeper · 2026-07-02T14:55:13Z

🦞🧹
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.

hugenshen · 2026-07-05T02:07:51Z

@clawsweeper re-review

clawsweeper · 2026-07-05T02:07:54Z

🦞🧹
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.

sallyom · 2026-07-06T16:54:12Z

Merge-ready from local maintainer review. Autoreview is clean at head SHA 63570bb, CI is green, and I found no compatibility concerns or breaking changes. This is the best narrow fix for the remaining Google OAuth JSON parse-boundary gap: it uses the existing 16 MiB provider JSON cap at the owning call sites while preserving normal OAuth behavior.

Co-authored-by: NIO <[email protected]>

openclaw-barnacle Bot added extensions: google size: XS labels Jun 28, 2026

openclaw-barnacle Bot added size: S and removed size: XS labels Jun 28, 2026

clawsweeper Bot added the merge-risk: 🚨 auth-provider 🚨 May break OAuth, tokens, provider routing, model choice, or credentials. label Jun 28, 2026

hugenshen force-pushed the fix/bound-google-oauth-json-responses branch from 1a7d1ff to 0bb0e96 Compare June 29, 2026 00:45

openclaw-barnacle Bot added the scripts Repository scripts label Jun 29, 2026

clawsweeper Bot added merge-risk: 🚨 automation 🚨 May affect CI, automerge, proof capture, label sync, or maintainer automation. and removed merge-risk: 🚨 automation 🚨 May affect CI, automerge, proof capture, label sync, or maintainer automation. labels Jun 29, 2026

wangmiao0668000666 mentioned this pull request Jun 29, 2026

fix(google): bound google OAuth fetchWithTimeout arrayBuffer at 16 MiB #97628

Merged

This was referenced Jun 29, 2026

fix(google): bound OAuth token response reads to prevent OOM #97703

Closed

fix(openrouter,google): bound OAuth response body reads #97745

Closed

fix(google): bound OAuth token exchange response reads to prevent OOM #98223

Closed

This was referenced Jul 1, 2026

fix(tlon): bound error response body reads to prevent OOM #98496

Merged

fix(openai): bound OAuth and image-generation error body reads to prevent OOM #98502

Closed

This was referenced Jul 1, 2026

fix(google): bound OAuth token error body reads to prevent OOM #98709

Closed

fix(google): bound OAuth token error body reads to prevent OOM #98722

Closed

hugenshen force-pushed the fix/bound-google-oauth-json-responses branch from 0bb0e96 to 09fef5a Compare July 2, 2026 14:47

openclaw-barnacle Bot removed the scripts Repository scripts label Jul 2, 2026

clawsweeper Bot mentioned this pull request Jul 3, 2026

fix(google): bound OAuth token error response reads #99605

Merged

hugenshen force-pushed the fix/bound-google-oauth-json-responses branch from 09fef5a to d7da56d Compare July 4, 2026 14:03

fix(google): bound OAuth JSON response reads

63570bb

hugenshen force-pushed the fix/bound-google-oauth-json-responses branch from d7da56d to 63570bb Compare July 4, 2026 16:23

sallyom self-assigned this Jul 6, 2026

sallyom merged commit 2e967ea into openclaw:main Jul 6, 2026
88 checks passed

This was referenced Jul 6, 2026

fix(extensions/openai): bound ChatGPT OAuth token response reads to prevent OOM #101075

Closed

fix(msteams): bound Graph attachment JSON responses #101082

Merged

github-actions Bot pushed a commit to Desicool/openclaw that referenced this pull request Jul 7, 2026

fix(google): bound OAuth JSON response reads (openclaw#97587)

f805c9a

Co-authored-by: NIO <[email protected]>

giodl73-repo pushed a commit to giodl73-repo/openclaw that referenced this pull request Jul 8, 2026

fix(google): bound OAuth JSON response reads (openclaw#97587)

2cc972b

Co-authored-by: NIO <[email protected]>

clawsweeper Bot mentioned this pull request Jul 17, 2026

fix(google-oauth): cancel unread response body on non-ok HTTP responses #109975

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Uh oh!

fix(google): bound OAuth project and token JSON response reads#97587

fix(google): bound OAuth project and token JSON response reads#97587
sallyom merged 1 commit into
openclaw:mainfrom
hugenshen:fix/bound-google-oauth-json-responses

hugenshen commented Jun 28, 2026 •

edited

Loading

Uh oh!

clawsweeper Bot commented Jun 28, 2026 •

edited

Loading

Uh oh!

hugenshen commented Jul 2, 2026

Uh oh!

clawsweeper Bot commented Jul 2, 2026 •

edited

Loading

Uh oh!

hugenshen commented Jul 5, 2026

Uh oh!

clawsweeper Bot commented Jul 5, 2026 •

edited

Loading

Uh oh!

sallyom commented Jul 6, 2026 •

edited

Loading

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Uh oh!

Uh oh!

Conversation

hugenshen commented Jun 28, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

What Problem This Solves

Changes

Real behavior proof

Evidence

Uh oh!

clawsweeper Bot commented Jun 28, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

hugenshen commented Jul 2, 2026

Uh oh!

clawsweeper Bot commented Jul 2, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

hugenshen commented Jul 5, 2026

Uh oh!

clawsweeper Bot commented Jul 5, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

sallyom commented Jul 6, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

hugenshen commented Jun 28, 2026 •

edited

Loading

clawsweeper Bot commented Jun 28, 2026 •

edited

Loading

clawsweeper Bot commented Jul 2, 2026 •

edited

Loading

clawsweeper Bot commented Jul 5, 2026 •

edited

Loading

sallyom commented Jul 6, 2026 •

edited

Loading