chore(effect): bump published provider/sandbox pins to 0.4.15/0.4.18

[ThomasK33]

Summary

Move the private Effect bridge to the freshly published Coder releases. No source logic or published-package files change.

Exact pin	Before	After
@coder/ai-sdk-provider	0.4.14	0.4.15
@coder/ai-sdk-sandbox	0.4.17	0.4.18

1. Update the two exact pins in packages/effect/package.json and their lockfile entries.
2. Replace the two version-specific minimumReleaseAgeExclude entries with @coder/ai-sdk-provider@0.4.15 and @coder/ai-sdk-sandbox@0.4.18. These newly published releases need the exclusions.
3. Update the two README “Pinned surface” rows and run pnpm format.

@ai-sdk/provider stays at 4.0.14: npm view @coder/ai-sdk-provider@0.4.15 dependencies.@ai-sdk/provider returns 4.0.14, so parity already holds.

packages/effect remains private: true and is outside release-please. This change should not create a release PR.

Validation

Validated tree committed as edb19cabf89cee33e7e0dcbd1c4b01049e0d106f with Node 26.8.2 and pnpm 11.27.0.

Dependency check	Result
pnpm why -r @ai-sdk/provider	One version: 4.0.14
pnpm why -r zod	One version: 4.6.5
pnpm why -r @ai-sdk/mcp	One version: 2.0.49, unchanged

• pnpm install without CI=true: passed. The first invocation rejected the outgoing sandbox 0.4.17 lock entry, still under 24 hours old. The retry added that outgoing pin only to the command-line exclusion list; no extra exclusion is committed.
• pnpm dedupe --check: passed. No dedupe rewrite needed. The inspected lockfile diff contains only the two Coder release replacements and their integrity values.
• CI=true pnpm install --frozen-lockfile: passed with the committed exclusions only.
• pnpm check: passed (format, lint, and typecheck).
• pnpm -r build: passed.
• pnpm -r test: 640 passed — agent 369, sandbox 191, effect 45, provider 22, release-please-ai 13.
• pnpm publint and pnpm attw: passed.
• Live e2e on this head: 7/7 passed, 2026-09-15T22:05:27Z–22:06:05Z, at https://dogfood.cdr.dev; server v2.37.1-devel+5699a5cf04. coder whoami verified ambient authentication. Both Vitest and tee exited 0.

Verification tiers: dependency and release-scope claims verified against registry metadata/source; agent behavior verified live against dogfood.

Live dogfood evidence

Executed from packages/agent:

npx vitest run test/e2e 2>&1 | tee ~/maintainer-scratch/w35-e2e-20260915T220527Z.log

The full output is retained below (ANSI formatting removed only). The video is an accelerated replay of the captured terminal output, not real-time playback or a second test run. The 36.98-second test run completed at 22:06:05Z; the terminal wrapper later timed out waiting for its shell marker, after the test and tee had already exited 0.

live-e2e.webm

 RUN  v5.0.0 /home/coder/.mux/src/ai-sdk/deps-effect-pins-sep15/packages/agent

stdout | test/e2e/agent.e2e.test.ts > CoderAgent e2e (live Coder) > emits a coherent transport-event trace for a client-tool turn (#45)
[+     0ms] segment:start 
[+     0ms] http:request GET /api/v2/organizations/703f72a1-76f6-4f89-9de6-8a3989693fe5/chats/models
[+   374ms] http:response GET /api/v2/organizations/703f72a1-76f6-4f89-9de6-8a3989693fe5/chats/models → 200 (373ms)
[+   487ms] http:request POST /api/v2/chats
[+   698ms] http:response POST /api/v2/chats → 201 (211ms)
[+   699ms] ws:dial attempt=1
[+  1101ms] ws:open 
[+  1111ms] ws:event message
[+  1111ms] ws:event status:running
[+  1111ms] ws:event preview_reset
[+  6098ms] ws:event message_part
[+  6304ms] ws:event message_part
[+  6305ms] ws:event message_part
[+  6305ms] ws:event message_part
[+  6305ms] ws:event message_part
[+  6343ms] ws:event message
[+  6343ms] ws:event preview_reset
[+  6393ms] ws:event status:requires_action
[+  6393ms] ws:event action_required
[+  6393ms] segment:settle status=requires_action finish=tool-calls (6393ms)
[+  6396ms] segment:start 
[+  6396ms] http:request POST /api/v2/chats/d1591afb-6d47-42c6-b485-a64dca438c5d/tool-results
[+  6771ms] http:response POST /api/v2/chats/d1591afb-6d47-42c6-b485-a64dca438c5d/tool-results → 204 (375ms)
[+  6777ms] ws:event message
[+  6777ms] ws:event status:running
[+  6777ms] ws:event preview_reset
[+  6837ms] ws:event preview_reset
[+  8131ms] ws:event message_part
[+  8131ms] ws:event message_part
[+  8187ms] ws:event message
[+  8187ms] ws:event preview_reset
[+  8232ms] ws:event status:waiting
[+  8232ms] ws:close 
[+  8232ms] segment:settle status=waiting finish=stop (1837ms)

 ✓ test/e2e/agent.e2e.test.ts (7 tests) 36550ms
   ✓ CoderAgent e2e (live Coder) (7)
     ✓ generates plain text 2137ms
     ✓ streams text deltas 3307ms
     ✓ round-trips a custom (client-executed) tool 6503ms
     ✓ reuses ONE WebSocket across a multi-tool-step turn (#44) 12060ms
     ✓ emits a coherent transport-event trace for a client-tool turn (#45) 8384ms
     ✓ uploads and downloads a chat file (round-trip) 481ms
     ✓ attaches a file and the model reads its contents 3225ms

 Test Files  1 passed (1)
      Tests  7 passed (7)
   Start at  22:05:28
   Duration  36.98s (tests 99%, transform 1%)

---

Generated with xum • Model: coder:openai/gpt-6-astra • Thinking: xhigh

[ThomasK33]

@codex review

[ThomasK33]

@codex security review

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-09-15T22:15:14.672546Z	edb19ca	Manual request
🔒 Security Review	✅ Completed 2026-09-15T22:16:20.972450Z	edb19ca	Manual request

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

[chatgpt-codex-connector]

Codex Review: Didn't find any major issues. Already looking forward to the next diff.

Reviewed commit: edb19cabf8

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

🛡️ Codex Security Review

Security review completed. No security issues were found in this pull request.

Reviewed commit: edb19cabf8

View security finding report

Only the user who started this review can view the report in Codex.

ℹ️ About Codex security reviews in GitHub

This is an experimental Codex feature. Security reviews are triggered when:

• You comment "@codex security review"
• A regular code review gets triggered (for example, "@codex review" or when a PR is opened), and you’re opted in so security review runs alongside code review

Once complete, Codex will leave suggestions, or a comment if no findings are found.