Follow-up to #2344 (NemoClaw 0.0.24). The http.request → https.request
rewrite that resolves the NODE_USE_ENV_PROXY=1 / library-side proxy
double-processing was shallow-copying the caller's options into the
rewritten https.request. That dragged three classes of forward-proxy-
hop fields onto a request that now goes direct to the upstream:

  * options.agent — a forward-proxy http.Agent. http.Agent cannot speak
    TLS, so https.request either ignores it and falls back to a default
    agent or fails the TLS handshake outright depending on caller. This
    is the most likely root cause of the deepinfra-style report on
    Discord ("LLM request failed: network connection error" against a
    custom OpenAI-compatible upstream while NVIDIA-routed providers
    keep working). NVIDIA Endpoints' DNS-rewritten path through OpenShell
    doesn't end up in this branch, so the bug stayed invisible.

  * options.auth — basic-auth meant for the proxy hop. Leaving it on
    https.request would Basic-auth the *target* server with proxy
    credentials.

  * Host / Proxy-Authorization / Proxy-Connection / Proxy-Authenticate
    headers — Host points at the proxy host and the Proxy-* headers are
    hop-by-hop. Stripping them lets Node regenerate Host from the
    target's hostname/port and prevents proxy creds from leaking
    upstream.

Adds test/http-proxy-fix-rewrite.test.ts with seven cases pinning each
strip and confirming signal/timeout/TLS fields and target headers
(Authorization, Content-Type, …) survive. The existing
http-proxy-fix-sync.test.ts byte-for-byte enforcer still passes — the
canonical wrapper at nemoclaw-blueprint/scripts/http-proxy-fix.js and
the heredoc in scripts/nemoclaw-start.sh were updated together.

Reverts the openclaw `request.allowPrivateNetwork` Dockerfile change
that this PR previously carried — empirical investigation showed
openclaw 2026.4.9 rejects that key in strict zod and `openclaw doctor
--fix` (which the Dockerfile runs at build time) silently strips it
back to `"request": {}` before the image ships. The upstream plumbing
that reads `request.allowPrivateNetwork` only landed in 2026.4.10
(commit 0808dd111c, openclaw PR #63671). The change here was a no-op,
diagnosing the wrong layer; this commit replaces it with the actual
fix at the wrapper layer.

The label-rename + arithmetic-via-`--json` test improvements added
earlier on this branch (Phase 4c, TC-SBX-02 rewrite, skill-verify
hardening, sandbox_exec stderr-merge fix) are kept — they catch
regressions independent of the proxy bug.

Signed-off-by: Aaron Erickson <aerickson@nvidia.com>

…est (#2490)

## Summary

Follow-up to #2344 (NemoClaw 0.0.24). The `http.request` →
`https.request` rewrite in
`nemoclaw-blueprint/scripts/http-proxy-fix.js` was shallow-copying the
caller's options into the rewritten `https.request` call. That dragged
forward-proxy-hop fields onto a request that now goes direct to the
upstream. This is consistent with the Discord report from `mflova` after
the 0.0.23 → 0.0.24 bump:

> Finally, when I run the bot, I always get an error after sending any
message to it: `error=LLM request failed: network connection error.
rawError=Connection error`
>
> Some context: I am using 0.0.23 as the 0.0.24 had a bug with the
network.

mflova was on a custom OpenAI-compatible endpoint (deepinfra). NVIDIA
Endpoints' DNS-rewritten path through OpenShell doesn't end up in the
wrapper's FORWARD-mode branch, which is why upstream regressions there
stayed invisible to nightly-e2e. mflova's reported error is the openclaw
client's wrapped surface; on Node 22 the underlying mechanism is a
synchronous `TypeError: Protocol "https:" not supported. Expected
"http:"` thrown by `https.request` when the forward-proxy `http.Agent`
reaches it. On Node 18/20 the same root cause manifests as a TLS
handshake failure rather than a synchronous throw — same bug class,
different surface.

## Bisect proof (committed in this PR)

`test/http-proxy-fix-rewrite.test.ts` carries both halves of the proof
in-repo:

- The "control" `describe` block reproduces the *broken* pre-fix rewrite
shape inline (`Object.assign` with no field strips, hop-by-hop headers
preserved) and asserts `https.request` rejects it with `TypeError:
Protocol "https:" not supported`. That test passes regardless of the
wrapper, locking in the bug class.
- The rewrite `describe` block asserts the wrapper produces options that
*don't* match that shape — `agent`, `auth`, `Host`, `Proxy-*`, the rest
of RFC 7230 §6.1 hop-by-hop, `servername`, `checkServerIdentity`,
`socketPath`, `localAddress`, `lookup`, `family`, `hints` are all
stripped on rewrite.

Combined: bug class is real and detectable; wrapper's job is to never
produce that shape. If a future maintainer reverts a strip, the rewrite
test breaks; if Node changes the failure surface, the control test
breaks. Two-sided coverage without storing a copy of the broken wrapper.

## What was wrong

The rewrite was:

```js
var rewritten = Object.assign({}, options, {
  hostname: target.hostname, host: target.hostname,
  port: target.port || 443, path: target.pathname + target.search,
  protocol: 'https:',
});
return https.request(rewritten, callback);
```

Three classes of forward-proxy-hop fields rode along:

- **`options.agent`** — a forward-proxy `http.Agent`.
`http.Agent.protocol === 'http:'`, so on Node 22 `https.request` throws
`TypeError: Protocol "https:" not supported. Expected "http:"`
synchronously; on Node 18/20 it falls through and the TLS handshake
fails. Manifests upstream as "Connection error".
- **`options.auth`** — basic-auth meant for the proxy hop. Leaving it on
`https.request` Basic-auths the *target* server with proxy credentials.
- **Hop-by-hop headers (RFC 7230 §6.1)** — `Connection`, `Keep-Alive`,
`Proxy-Authorization`, `TE`, `Trailer`, `Transfer-Encoding`, `Upgrade`,
plus tokens named in `Connection` (transitively hop-by-hop), plus `Host`
(proxy-pointing) and `Proxy-Connection` (de facto deprecated).
`Connection: close` from the proxy hop defeats keep-alive on the
rewrite; `Upgrade` to a non-WS target produces 400/426; `Proxy-*` leak
proxy credentials upstream.

Plus: TLS identity (`servername`, `checkServerIdentity`), socket-path
proxy hint (`socketPath`), and source-binding / DNS hints
(`localAddress`, `lookup`, `family`, `hints`) that all describe the
connection to the proxy and don't apply to a different upstream.

## The fix

`nemoclaw-blueprint/scripts/http-proxy-fix.js` (canonical) and the
heredoc in `scripts/nemoclaw-start.sh` (byte-for-byte synced via the
existing `http-proxy-fix-sync.test.ts`):

- `sanitizeHeaders()` strips the full RFC 7230 §6.1 hop-by-hop set plus
`Host`, `Proxy-Connection`, `Proxy-Authenticate`, and any token named in
the `Connection` header (transitive per the same RFC). Case-insensitive.
- `delete rewritten.agent`, `delete rewritten.auth`, `delete
rewritten.servername`, `delete rewritten.checkServerIdentity`, `delete
rewritten.socketPath`, `delete rewritten.localAddress`, `delete
rewritten.lookup`, `delete rewritten.family`, `delete rewritten.hints`
after the `Object.assign`.
- Signal (AbortController), TLS material
(`ca`/`cert`/`key`/`rejectUnauthorized`), `timeout`, body, and
target-intent headers (`Authorization`, `Content-Type`, …) all survive.

## Tests

- **`test/http-proxy-fix-rewrite.test.ts` (12 cases)** — spy-level
rewrite tests pinning every strip; control test reproducing the
broken-wrapper shape and asserting `TypeError`.
- **`test/http-proxy-fix-e2e.test.ts` (1 case)** — end-to-end. openssl
shell-out at module load (skipped locally if openssl missing; loud-fails
under `CI=true`). Spins up an HTTPS mock on `127.0.0.1:0`, builds the
FORWARD-mode `http.request` shape with forward-proxy `http.Agent` +
proxy basic-auth + proxy-pointing Host, asserts the round trip
completes.
- **`test/http-proxy-fix-sync.test.ts` (6 cases, existing)** —
byte-for-byte parity between the canonical wrapper and the
`scripts/nemoclaw-start.sh` heredoc still enforced.

## What this PR no longer carries

This branch previously included a Dockerfile change that added
`'request': {'allowPrivateNetwork': True}` to the inference provider in
the baked `openclaw.json`, and a regression test for it. Empirical
investigation showed:

- openclaw 2026.4.9's strict zod schema rejects
`request.allowPrivateNetwork` as an unrecognized key
(`src/config/zod-schema.core.ts:283-292` in v2026.4.9).
- `openclaw doctor --fix` — which the Dockerfile runs at image build
time — silently strips the key, leaving `"request": {}` before the image
ships.
- The plumbing that *reads* `request.allowPrivateNetwork` only landed in
2026.4.10 (upstream PR #63671, commit `0808dd111c`).

So the previous "fix" was a no-op diagnosing the wrong layer. Dockerfile
and `test/sandbox-provisioning.test.ts` are restored to `origin/main`.

## What this PR keeps from the prior iteration

The label-rename + arithmetic-via-`--json` test improvements added
earlier on this branch are kept — they catch regressions independent of
the proxy bug:

- `test/e2e/test-full-e2e.sh`: Phase 4b relabelled `[LIVE]` →
`[ROUTING]`; new Phase 4c that runs `openclaw agent --json` over SSH
inside the sandbox and asserts the model answered `42` to "What is 6
multiplied by 7?". Phase 4c is the only assertion in the suite that
exercises the openclaw HTTP client end-to-end against `inference.local`.
- `test/e2e/test-hermes-e2e.sh`: same `[LIVE]` → `[ROUTING]` relabel.
- `test/e2e/test-sandbox-operations.sh` TC-SBX-02: replaces `Say
exactly: HELLO_E2E` + merged-output grep with the
arithmetic-via-`--json` pattern.
-
`test/e2e/e2e-cloud-experimental/features/skill/verify-sandbox-skill-via-agent.sh`:
stops embedding `${VERIFY_TOKEN}` in the prompt; refuses overrides that
smuggle it back; adds a negative assertion on `SsrFBlockedError` /
transport / gateway-unavailable markers.

## Type of Change

- [x] Code change (feature, bug fix, or refactor)

## Verification

- [x] `npx vitest run test/http-proxy-fix-e2e.test.ts
test/http-proxy-fix-rewrite.test.ts test/http-proxy-fix-sync.test.ts
test/sandbox-provisioning.test.ts` — 28/28 pass.
- [x] Adversarial review's blockers (B1–B3) and concerns (C1–C7)
addressed in commit `69115de7`.
- [x] `bash -n` and `shellcheck -S warning` clean on
`scripts/nemoclaw-start.sh` and the four touched e2e scripts.
- [x] Heredoc-sync test still passes; canonical wrapper and
`scripts/nemoclaw-start.sh` heredoc updated together.
- [x] No secrets, API keys, or credentials committed.

## Per-commit on this branch

- `f162daa5` — initial strip of `agent`/`auth`/`Host`/`Proxy-*` (the
fix), spy-level unit tests
- `2f217f8a` — end-to-end test against a local HTTPS mock with
self-signed cert
- `69115de7` — wider RFC 7230 §6.1 hop-by-hop + TLS/transport hint
strips, in-repo bisect control, vi.stubEnv consistency, http.request
restore, 7-day cert, CI strict-skip

## How to validate after merge

The signal job is `cloud-e2e` in `nightly-e2e.yaml` running
`test/e2e/test-full-e2e.sh`. Watch for:

```
PASS: [LIVE] openclaw agent: model answered 6×7=42 through openclaw → inference.local
```

For the deepinfra-style failure specifically, the unit + e2e + control
tests in this PR pin every strip the wrapper performs *and* the bug
class it prevents. A real proxy + real upstream + real TLS round trip
would need a CI job with a non-NVIDIA OpenAI-compatible API key — its
own follow-up PR if you want belt-and-suspenders coverage.

<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->

## Summary by CodeRabbit

## Release Notes

* **Bug Fixes**
* Improved HTTP proxy request handling to sanitize headers and remove
incompatible routing-specific fields, ensuring proper downstream
forwarding.

* **Tests**
* Added comprehensive test coverage for proxy header sanitization and
rewrite behavior.
* Enhanced E2E testing with improved routing-layer assertions and JSON
payload extraction from agent responses.

<!-- end of auto-generated comment: release notes by coderabbit.ai -->

---------

Signed-off-by: Aaron Erickson <aerickson@nvidia.com>

…est (NVIDIA#2490)

## Summary

Follow-up to NVIDIA#2344 (NemoClaw 0.0.24). The `http.request` →
`https.request` rewrite in
`nemoclaw-blueprint/scripts/http-proxy-fix.js` was shallow-copying the
caller's options into the rewritten `https.request` call. That dragged
forward-proxy-hop fields onto a request that now goes direct to the
upstream. This is consistent with the Discord report from `mflova` after
the 0.0.23 → 0.0.24 bump:

> Finally, when I run the bot, I always get an error after sending any
message to it: `error=LLM request failed: network connection error.
rawError=Connection error`
>
> Some context: I am using 0.0.23 as the 0.0.24 had a bug with the
network.

mflova was on a custom OpenAI-compatible endpoint (deepinfra). NVIDIA
Endpoints' DNS-rewritten path through OpenShell doesn't end up in the
wrapper's FORWARD-mode branch, which is why upstream regressions there
stayed invisible to nightly-e2e. mflova's reported error is the openclaw
client's wrapped surface; on Node 22 the underlying mechanism is a
synchronous `TypeError: Protocol "https:" not supported. Expected
"http:"` thrown by `https.request` when the forward-proxy `http.Agent`
reaches it. On Node 18/20 the same root cause manifests as a TLS
handshake failure rather than a synchronous throw — same bug class,
different surface.

## Bisect proof (committed in this PR)

`test/http-proxy-fix-rewrite.test.ts` carries both halves of the proof
in-repo:

- The "control" `describe` block reproduces the *broken* pre-fix rewrite
shape inline (`Object.assign` with no field strips, hop-by-hop headers
preserved) and asserts `https.request` rejects it with `TypeError:
Protocol "https:" not supported`. That test passes regardless of the
wrapper, locking in the bug class.
- The rewrite `describe` block asserts the wrapper produces options that
*don't* match that shape — `agent`, `auth`, `Host`, `Proxy-*`, the rest
of RFC 7230 §6.1 hop-by-hop, `servername`, `checkServerIdentity`,
`socketPath`, `localAddress`, `lookup`, `family`, `hints` are all
stripped on rewrite.

Combined: bug class is real and detectable; wrapper's job is to never
produce that shape. If a future maintainer reverts a strip, the rewrite
test breaks; if Node changes the failure surface, the control test
breaks. Two-sided coverage without storing a copy of the broken wrapper.

## What was wrong

The rewrite was:

```js
var rewritten = Object.assign({}, options, {
  hostname: target.hostname, host: target.hostname,
  port: target.port || 443, path: target.pathname + target.search,
  protocol: 'https:',
});
return https.request(rewritten, callback);
```

Three classes of forward-proxy-hop fields rode along:

- **`options.agent`** — a forward-proxy `http.Agent`.
`http.Agent.protocol === 'http:'`, so on Node 22 `https.request` throws
`TypeError: Protocol "https:" not supported. Expected "http:"`
synchronously; on Node 18/20 it falls through and the TLS handshake
fails. Manifests upstream as "Connection error".
- **`options.auth`** — basic-auth meant for the proxy hop. Leaving it on
`https.request` Basic-auths the *target* server with proxy credentials.
- **Hop-by-hop headers (RFC 7230 §6.1)** — `Connection`, `Keep-Alive`,
`Proxy-Authorization`, `TE`, `Trailer`, `Transfer-Encoding`, `Upgrade`,
plus tokens named in `Connection` (transitively hop-by-hop), plus `Host`
(proxy-pointing) and `Proxy-Connection` (de facto deprecated).
`Connection: close` from the proxy hop defeats keep-alive on the
rewrite; `Upgrade` to a non-WS target produces 400/426; `Proxy-*` leak
proxy credentials upstream.

Plus: TLS identity (`servername`, `checkServerIdentity`), socket-path
proxy hint (`socketPath`), and source-binding / DNS hints
(`localAddress`, `lookup`, `family`, `hints`) that all describe the
connection to the proxy and don't apply to a different upstream.

## The fix

`nemoclaw-blueprint/scripts/http-proxy-fix.js` (canonical) and the
heredoc in `scripts/nemoclaw-start.sh` (byte-for-byte synced via the
existing `http-proxy-fix-sync.test.ts`):

- `sanitizeHeaders()` strips the full RFC 7230 §6.1 hop-by-hop set plus
`Host`, `Proxy-Connection`, `Proxy-Authenticate`, and any token named in
the `Connection` header (transitive per the same RFC). Case-insensitive.
- `delete rewritten.agent`, `delete rewritten.auth`, `delete
rewritten.servername`, `delete rewritten.checkServerIdentity`, `delete
rewritten.socketPath`, `delete rewritten.localAddress`, `delete
rewritten.lookup`, `delete rewritten.family`, `delete rewritten.hints`
after the `Object.assign`.
- Signal (AbortController), TLS material
(`ca`/`cert`/`key`/`rejectUnauthorized`), `timeout`, body, and
target-intent headers (`Authorization`, `Content-Type`, …) all survive.

## Tests

- **`test/http-proxy-fix-rewrite.test.ts` (12 cases)** — spy-level
rewrite tests pinning every strip; control test reproducing the
broken-wrapper shape and asserting `TypeError`.
- **`test/http-proxy-fix-e2e.test.ts` (1 case)** — end-to-end. openssl
shell-out at module load (skipped locally if openssl missing; loud-fails
under `CI=true`). Spins up an HTTPS mock on `127.0.0.1:0`, builds the
FORWARD-mode `http.request` shape with forward-proxy `http.Agent` +
proxy basic-auth + proxy-pointing Host, asserts the round trip
completes.
- **`test/http-proxy-fix-sync.test.ts` (6 cases, existing)** —
byte-for-byte parity between the canonical wrapper and the
`scripts/nemoclaw-start.sh` heredoc still enforced.

## What this PR no longer carries

This branch previously included a Dockerfile change that added
`'request': {'allowPrivateNetwork': True}` to the inference provider in
the baked `openclaw.json`, and a regression test for it. Empirical
investigation showed:

- openclaw 2026.4.9's strict zod schema rejects
`request.allowPrivateNetwork` as an unrecognized key
(`src/config/zod-schema.core.ts:283-292` in v2026.4.9).
- `openclaw doctor --fix` — which the Dockerfile runs at image build
time — silently strips the key, leaving `"request": {}` before the image
ships.
- The plumbing that *reads* `request.allowPrivateNetwork` only landed in
2026.4.10 (upstream PR #63671, commit `0808dd111c`).

So the previous "fix" was a no-op diagnosing the wrong layer. Dockerfile
and `test/sandbox-provisioning.test.ts` are restored to `origin/main`.

## What this PR keeps from the prior iteration

The label-rename + arithmetic-via-`--json` test improvements added
earlier on this branch are kept — they catch regressions independent of
the proxy bug:

- `test/e2e/test-full-e2e.sh`: Phase 4b relabelled `[LIVE]` →
`[ROUTING]`; new Phase 4c that runs `openclaw agent --json` over SSH
inside the sandbox and asserts the model answered `42` to "What is 6
multiplied by 7?". Phase 4c is the only assertion in the suite that
exercises the openclaw HTTP client end-to-end against `inference.local`.
- `test/e2e/test-hermes-e2e.sh`: same `[LIVE]` → `[ROUTING]` relabel.
- `test/e2e/test-sandbox-operations.sh` TC-SBX-02: replaces `Say
exactly: HELLO_E2E` + merged-output grep with the
arithmetic-via-`--json` pattern.
-
`test/e2e/e2e-cloud-experimental/features/skill/verify-sandbox-skill-via-agent.sh`:
stops embedding `${VERIFY_TOKEN}` in the prompt; refuses overrides that
smuggle it back; adds a negative assertion on `SsrFBlockedError` /
transport / gateway-unavailable markers.

## Type of Change

- [x] Code change (feature, bug fix, or refactor)

## Verification

- [x] `npx vitest run test/http-proxy-fix-e2e.test.ts
test/http-proxy-fix-rewrite.test.ts test/http-proxy-fix-sync.test.ts
test/sandbox-provisioning.test.ts` — 28/28 pass.
- [x] Adversarial review's blockers (B1–B3) and concerns (C1–C7)
addressed in commit `69115de7`.
- [x] `bash -n` and `shellcheck -S warning` clean on
`scripts/nemoclaw-start.sh` and the four touched e2e scripts.
- [x] Heredoc-sync test still passes; canonical wrapper and
`scripts/nemoclaw-start.sh` heredoc updated together.
- [x] No secrets, API keys, or credentials committed.

## Per-commit on this branch

- `f162daa5` — initial strip of `agent`/`auth`/`Host`/`Proxy-*` (the
fix), spy-level unit tests
- `2f217f8a` — end-to-end test against a local HTTPS mock with
self-signed cert
- `69115de7` — wider RFC 7230 §6.1 hop-by-hop + TLS/transport hint
strips, in-repo bisect control, vi.stubEnv consistency, http.request
restore, 7-day cert, CI strict-skip

## How to validate after merge

The signal job is `cloud-e2e` in `nightly-e2e.yaml` running
`test/e2e/test-full-e2e.sh`. Watch for:

```
PASS: [LIVE] openclaw agent: model answered 6×7=42 through openclaw → inference.local
```

For the deepinfra-style failure specifically, the unit + e2e + control
tests in this PR pin every strip the wrapper performs *and* the bug
class it prevents. A real proxy + real upstream + real TLS round trip
would need a CI job with a non-NVIDIA OpenAI-compatible API key — its
own follow-up PR if you want belt-and-suspenders coverage.

<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->

## Summary by CodeRabbit

## Release Notes

* **Bug Fixes**
* Improved HTTP proxy request handling to sanitize headers and remove
incompatible routing-specific fields, ensuring proper downstream
forwarding.

* **Tests**
* Added comprehensive test coverage for proxy header sanitization and
rewrite behavior.
* Enhanced E2E testing with improved routing-layer assertions and JSON
payload extraction from agent responses.

<!-- end of auto-generated comment: release notes by coderabbit.ai -->

---------

Signed-off-by: Aaron Erickson <aerickson@nvidia.com>

Follow-up to NVIDIA#2344 (the http.request wrapper for NVIDIA#2109). PR NVIDIA#2344
covers axios / follow-redirects / proxy-from-env, all of which flow
through Node's http.request and EnvHttpProxyAgent. This change covers
an additional code path the Microsoft Teams adapter uses: native
fetch() with a custom undici dispatcher.

When OpenClaw's Teams adapter handles a channel @mention it calls the
Microsoft Graph API via:

    fetch('https://graph.microsoft.com/v1.0/teams/.../messages/...',
          { dispatcher: customUndiciAgent })

The custom dispatcher routes the request through its own connection
pool, bypassing the default EnvHttpProxyAgent — which is the only
thing routing HTTPS through the OpenShell L7 proxy. Direct egress to
graph.microsoft.com is blocked by the sandbox network namespace and
surfaces as ECONNREFUSED. Channel @mentions silently fail; DMs work
because the webhook payload carries the message body and no Graph
call is needed.

Wrap globalThis.fetch in the same preload that wraps http.request.
When fetch() is called with a custom dispatcher and an HTTPS URL,
strip the dispatcher and let EnvHttpProxyAgent handle the request.
Non-HTTPS URLs and dispatcher-free calls pass through unchanged so
intra-sandbox HTTP traffic and any non-proxy use of the dispatcher
option keep working.

Refinements over the originally proposed fix:
  - URL extraction handles all three fetch() input forms — string,
    URL object (.href), Request object (.url) — in that order. A
    naive url?.url check would miss URL instances (which have .href,
    not .url) and silently leave the dispatcher attached, defeating
    the fix on callers that pass URL instances.
  - One-shot console.warn so the strip is auditable in logs without
    spamming on every call. The flag is closure-scoped so a process
    restart re-arms it.
  - Defensive typeof origFetch === function guard so a Node runtime
    that ever ships without globalThis.fetch silently no-ops instead
    of throwing at preload time.

After NVIDIA#3109 the preload is delivered as a standalone module copied
from /usr/local/lib/nemoclaw/preloads/ at boot, so this PR only
touches the canonical http-proxy-fix.js — no heredoc to keep in sync.
http-proxy-fix-sync.test.ts already runs the entrypoint block end-to-
end and reads the generated file; extended with explicit assertions
for http.request and globalThis.fetch wrapper presence so a future
deletion of either wrapper trips the test.

Verified end-to-end on a real proxy-enabled sandbox: bot replies to
Teams channel @mentions; DM regression check passes; non-Graph
fetch() and fetch() without a custom dispatcher unaffected.