fix(cli): replace config-set allow-list with mode-based new-path gate#2420
Merged
fix(cli): replace config-set allow-list with mode-based new-path gate#2420
Conversation
The #1973 fix stopped the silent stub-write, but the replacement isRecognizedConfigPath walked the currently-loaded config with hasOwnProperty at each segment — so any dotpath whose intermediate objects did not yet exist was rejected with the same message as a typo. That broke every documented per-agent override that targets an unset namespace, including the `provider.compatible-endpoint.timeout Seconds` workaround for the Ollama reasoning-stream timeout. Replace the walk with a top-level-namespace allow-list (provider, model, agent, agents, tools, mcpServers, runtime, version) plus syntactic guards for empty segments and prototype-pollution keys (__proto__, constructor, prototype, toString, hasOwnProperty). The gateway namespace is still blocked by its dedicated check at the call site. Surface the allow-list in the error message so users get a discoverable hint instead of trial-and-error. Tests updated to reflect the new semantics and gain regression coverage for first-time writes under an unset namespace (the #2400 scenario) and for prototype-pollution segments appearing anywhere in the path, not just at the root. Resolves #2400. Signed-off-by: Tinson Lai <tinsonl@nvidia.com>
Contributor
📝 WalkthroughWalkthroughconfigSet was made async and now validates dotpath syntax and safety (prototype-pollution, empty segments, numeric segments), prevents clobbering existing non-object ancestors, and gates creation of previously-unset keys via an opt-in flag/env or interactive confirmation; exports updated to surface the new validators and gate helpers, and CLI/tests/docs adjusted. Changes
Sequence Diagram(s)sequenceDiagram
participant CLI as Nemoclaw CLI
participant Module as sandbox-config
participant Config as Config file
participant Env as Env/CI
participant User as TTY/User
CLI->>Module: configSet(sandbox, { dotpath, value, acceptNewPath })
Module->>Module: validateConfigDotpath(dotpath)
alt invalid syntax / unsafe
Module-->>CLI: error(reason)
else valid syntax
Module->>Config: read existing value at dotpath
alt value exists
Module-->>Config: write new value
Module-->>CLI: success
else value missing
Module->>Env: check NEMOCLAW_CONFIG_ACCEPT_NEW_PATH or opts.acceptNewPath
alt accepted via flag/env
Module-->>Config: write new value
Module-->>CLI: success
else interactive
Module->>User: prompt confirmYesNo("create new path?")
User-->>Module: yes/no
alt yes
Module-->>Config: write new value
Module-->>CLI: success
else no
Module-->>CLI: abort with error
end
end
end
end
Estimated code review effort🎯 4 (Complex) | ⏱️ ~45 minutes Poem
🚥 Pre-merge checks | ✅ 4 | ❌ 1❌ Failed checks (1 warning)
✅ Passed checks (4 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. ✨ Finishing Touches📝 Generate docstrings
🧪 Generate unit tests (beta)
Comment |
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
Contributor
There was a problem hiding this comment.
Actionable comments posted: 1
🧹 Nitpick comments (1)
test/config-set.test.ts (1)
119-124: Add an explicit test for theprototypedenylist segment.
UNSAFE_KEY_SEGMENTSincludesprototype, but this specific case is not asserted here.Add the missing regression assertion
it("rejects prototype-pollution segments anywhere in the path", () => { expect(isRecognizedConfigPath("toString")).toBe(false); expect(isRecognizedConfigPath("agents.constructor")).toBe(false); + expect(isRecognizedConfigPath("agents.prototype.config")).toBe(false); expect(isRecognizedConfigPath("provider.__proto__.polluted")).toBe(false); expect(isRecognizedConfigPath("tools.hasOwnProperty")).toBe(false); });🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@test/config-set.test.ts` around lines 119 - 124, Add a regression assertion to the test "rejects prototype-pollution segments anywhere in the path" that explicitly checks for the denylisted segment "prototype": update the test in test/config-set.test.ts by adding an expect call using isRecognizedConfigPath with a path that includes "prototype" (for example "provider.prototype.polluted" or just "prototype") and assert it returns false; this ensures UNSAFE_KEY_SEGMENTS' "prototype" entry is covered alongside the existing checks for "toString", "constructor", "__proto__", and "hasOwnProperty".
🤖 Prompt for all review comments with AI agents
Verify each finding against the current code and only fix it if needed.
Inline comments:
In `@src/lib/sandbox-config.ts`:
- Around line 174-179: isRecognizedConfigPath currently treats any dotted path
that starts with a recognized top-level key as valid, allowing typos in nested
keys; update isRecognizedConfigPath to reject unknown deep paths by (1) keeping
the existing guard for falsy/unsafe segments (UNSAFE_KEY_SEGMENTS) and top-level
membership (RECOGNIZED_TOP_LEVEL_KEYS), and then (2) when keys.length > 1
validate the remainder of the path against an explicit whitelist/schema instead
of accepting any child — e.g. check the full dotpath against a
RECOGNIZED_CONFIG_PATHS set or walk a RECOGNIZED_SUBKEYS map keyed by parent
names to confirm each subsequent segment is an allowed child; ensure you
reference and preserve UNSAFE_KEY_SEGMENTS and RECOGNIZED_TOP_LEVEL_KEYS and
only return true when the full path is recognized by the chosen deep-path
whitelist/schema.
---
Nitpick comments:
In `@test/config-set.test.ts`:
- Around line 119-124: Add a regression assertion to the test "rejects
prototype-pollution segments anywhere in the path" that explicitly checks for
the denylisted segment "prototype": update the test in test/config-set.test.ts
by adding an expect call using isRecognizedConfigPath with a path that includes
"prototype" (for example "provider.prototype.polluted" or just "prototype") and
assert it returns false; this ensures UNSAFE_KEY_SEGMENTS' "prototype" entry is
covered alongside the existing checks for "toString", "constructor",
"__proto__", and "hasOwnProperty".
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Enterprise
Run ID: 9e8479f1-d88d-4bf7-aef2-7a81b02cb835
📒 Files selected for processing (2)
src/lib/sandbox-config.tstest/config-set.test.ts
Addresses the CodeRabbit note on #2420: the pure root allow-list was materially stricter than the underlying schema. If OpenClaw adds a new top-level namespace upstream before we update RECOGNIZED_TOP_LEVEL_KEYS here, users who already have it populated would see their next `config set` blocked even though the path is demonstrably valid. isRecognizedConfigPath now takes an optional config object. When the full dotpath already resolves in that config, we trust the path and bypass the allow-list — OpenClaw wrote it, so it is by definition recognised. First-time writes (the #2400 scenario) still fall through to the allow-list, and the syntactic + prototype-pollution guards still apply in both branches. Tests cover the three new cases: modify under an unlisted root accepted, unset under an unlisted root still rejected even with a config supplied, and prototype-pollution still rejected under every branch. Signed-off-by: Tinson Lai <tinsonl@nvidia.com>
Contributor
There was a problem hiding this comment.
♻️ Duplicate comments (1)
src/lib/sandbox-config.ts (1)
175-180:⚠️ Potential issue | 🟠 MajorRoot-only fallback still accepts unknown deep keys (typo protection remains weakened).
At Line 180, any nested path under an allow-listed root is accepted when it does not already exist. Example:
provider.compatible-endpoint.timeuotSeconds(typo) passes validation and gets written, which undermines typo detection for first-time writes.🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@src/lib/sandbox-config.ts` around lines 175 - 180, isRecognizedConfigPath currently permits any nested path under an allowed top-level key, which lets typos like provider.compatible-endpoint.timeuotSeconds pass; modify the fallback so that the top-level allowance only applies to single-segment dotpaths: keep the existing validations and extractDotpath check, but change the final return to require keys.length === 1 in addition to RECOGNIZED_TOP_LEVEL_KEYS.has(keys[0]) so only root-level names (no dots) are accepted by the fallback.
🤖 Prompt for all review comments with AI agents
Verify each finding against the current code and only fix it if needed.
Duplicate comments:
In `@src/lib/sandbox-config.ts`:
- Around line 175-180: isRecognizedConfigPath currently permits any nested path
under an allowed top-level key, which lets typos like
provider.compatible-endpoint.timeuotSeconds pass; modify the fallback so that
the top-level allowance only applies to single-segment dotpaths: keep the
existing validations and extractDotpath check, but change the final return to
require keys.length === 1 in addition to RECOGNIZED_TOP_LEVEL_KEYS.has(keys[0])
so only root-level names (no dots) are accepted by the fallback.
ℹ️ Review info
⚙️ Run configuration
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Enterprise
Run ID: 02dea3be-3812-4f54-aded-768a75730c3a
📒 Files selected for processing (2)
src/lib/sandbox-config.tstest/config-set.test.ts
…nset-keys Signed-off-by: Tinson Lai <tinsonl@nvidia.com>
The previous fix for #2400 introduced a top-level allow-list of config namespaces. CodeRabbit and the issue reporter both noted that any schema-aware allow-list will drift from upstream OpenClaw and cannot catch deep typos either way. Drop the validator entirely and gate first-time writes by mode instead: * Interactive (TTY, !NEMOCLAW_NON_INTERACTIVE) — print the diff and prompt y/N when the key path does not yet exist in the agent config. * Non-interactive — refuse with a clear message naming the override flag and env var; modify of an existing key still goes straight through. * --config-accept-new-path / NEMOCLAW_CONFIG_ACCEPT_NEW_PATH=1 — skip the prompt/refusal so scripts can opt in explicitly. Prototype-pollution and gateway-section guards stay in place via a small validateConfigDotpath helper that only checks dotpath syntax. Resolves #2400. Signed-off-by: Tinson Lai <tinsonl@nvidia.com>
laitingsheng
changed the title
Contributor
There was a problem hiding this comment.
Actionable comments posted: 2
🤖 Prompt for all review comments with AI agents
Verify each finding against the current code and only fix it if needed.
Inline comments:
In `@src/lib/sandbox-config.ts`:
- Around line 162-173: validateConfigDotpath currently allows numeric segments
which extractDotpath interprets as array indices while setDotpath always
constructs plain objects; to avoid corrupting arrays, update
validateConfigDotpath to reject any dotpath that contains a purely numeric
segment (e.g., /^\d+$/) in addition to the existing checks. Keep references to
validateConfigDotpath, extractDotpath, setDotpath, and UNSAFE_KEY_SEGMENTS in
the message so the change is targeted: return { ok: false, reason: "numeric
segment not allowed" } for numeric segments until setDotpath is updated to
preserve/handle arrays.
- Around line 385-395: The interactive detection in sandbox-config.ts treats any
non-empty NEMOCLAW_NON_INTERACTIVE as true (interactive = !!process.stdin.isTTY
&& !process.env.NEMOCLAW_NON_INTERACTIVE), which deviates from the CLI
convention of only disabling prompts when NEMOCLAW_NON_INTERACTIVE === "1";
update the interactive check to mirror other commands by testing
process.env.NEMOCLAW_NON_INTERACTIVE === "1" (or its negation) so that only the
exact value "1" disables prompts, and keep the surrounding logic that uses
opts.acceptNewPath and prints the console errors for non-interactive flows
unchanged.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Enterprise
Run ID: 14f65eab-39ad-45c7-8076-9531df1283ee
📒 Files selected for processing (5)
.agents/skills/nemoclaw-user-reference/references/troubleshooting.mddocs/reference/troubleshooting.mdsrc/lib/sandbox-config.tssrc/nemoclaw.tstest/config-set.test.ts
✅ Files skipped from review due to trivial changes (2)
- docs/reference/troubleshooting.md
- .agents/skills/nemoclaw-user-reference/references/troubleshooting.md
* validateConfigDotpath now refuses numeric segments. extractDotpath reads array indices but setDotpath always materialises plain objects, so allowing them silently corrupts arrays. Helm-style array editing was never advertised, never tested for write intent, and never worked via the previous walk-based validator either; refuse cleanly until it is designed in as a feature of its own. * The first-time-write gate now treats only NEMOCLAW_NON_INTERACTIVE='1' as non-interactive, matching the dominant convention in src/nemoclaw.ts and src/lib/onboard.ts. * Surface the array-index limitation in the troubleshooting page. Signed-off-by: Tinson Lai <tinsonl@nvidia.com>
Contributor
There was a problem hiding this comment.
🧹 Nitpick comments (1)
src/lib/sandbox-config.ts (1)
391-417: Consider adding branch tests for the new-path gate.This block introduces important behavior modes (
--config-accept-new-path,NEMOCLAW_CONFIG_ACCEPT_NEW_PATH=1, interactive confirm, and non-interactive refusal). I’d add explicit tests for each mode to prevent regressions in CLI UX and script flows.🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@src/lib/sandbox-config.ts` around lines 391 - 417, Add automated branch tests covering the new-path gate around the oldValue === undefined block: assert behavior when opts.acceptNewPath === true (write allowed), when NEMOCLAW_CONFIG_ACCEPT_NEW_PATH="1" (write allowed), when interactive TTY with confirmYesNo returning true (write allowed) and false (process exits with "Aborted."), and when non-interactive TTY (process exits and prints the non-interactive refusal message). In tests mock/stub process.stdin.isTTY, set/unset process.env.NEMOCLAW_CONFIG_ACCEPT_NEW_PATH, stub confirmYesNo, and spy on process.exit and console.error to verify the correct branches in the code paths referencing opts.key, opts.acceptNewPath, confirmYesNo, and the process.env flag.
🤖 Prompt for all review comments with AI agents
Verify each finding against the current code and only fix it if needed.
Nitpick comments:
In `@src/lib/sandbox-config.ts`:
- Around line 391-417: Add automated branch tests covering the new-path gate
around the oldValue === undefined block: assert behavior when opts.acceptNewPath
=== true (write allowed), when NEMOCLAW_CONFIG_ACCEPT_NEW_PATH="1" (write
allowed), when interactive TTY with confirmYesNo returning true (write allowed)
and false (process exits with "Aborted."), and when non-interactive TTY (process
exits and prints the non-interactive refusal message). In tests mock/stub
process.stdin.isTTY, set/unset process.env.NEMOCLAW_CONFIG_ACCEPT_NEW_PATH, stub
confirmYesNo, and spy on process.exit and console.error to verify the correct
branches in the code paths referencing opts.key, opts.acceptNewPath,
confirmYesNo, and the process.env flag.
ℹ️ Review info
⚙️ Run configuration
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Enterprise
Run ID: 2b924b37-8fb0-4c4c-8700-5a1fbfd083b1
📒 Files selected for processing (4)
.agents/skills/nemoclaw-user-reference/references/troubleshooting.mddocs/reference/troubleshooting.mdsrc/lib/sandbox-config.tstest/config-set.test.ts
✅ Files skipped from review due to trivial changes (2)
- docs/reference/troubleshooting.md
- .agents/skills/nemoclaw-user-reference/references/troubleshooting.md
ericksoa
reviewed
Apr 25, 2026
Contributor
ericksoa
left a comment
ericksoa
left a comment
There was a problem hiding this comment.
Review
Good design choice decoupling from OpenClaw's schema — the old isRecognizedConfigPath was genuinely broken for documented workflows like provider.compatible-endpoint.timeoutSeconds, and maintaining a parallel allow-list would have been a drag. The mode-based gate is the right call.
CI is green (wsl-e2e pending). A few things I'd want addressed before merge:
Missing test coverage for the new-key gate
The tests cover validateConfigDotpath well, but the actual behavioral change — the oldValue === undefined branch in configSet — has no test coverage. These are the most important paths in the PR:
- Non-interactive mode refusing a new key without
--config-accept-new-path - Non-interactive mode accepting with
--config-accept-new-path - The
NEMOCLAW_CONFIG_ACCEPT_NEW_PATH=1env var path
I'd ask for at least the non-interactive rejection and flag override cases before merging.
Mixed import style in sandbox-config.ts
import * as readline from "readline"; // ES import (new, line 15)
const fs = require("fs"); // CJS require (existing, line 16)The entire rest of the file uses require(). The mixed style works because TypeScript rewrites it, but it's inconsistent. Either const readline = require("readline") to match the file, or a comment noting it's intentional for the TS migration path.
Dead code in confirmYesNo
if (!process.stdin.isTTY) {
if (typeof process.stdin.pause === "function") process.stdin.pause();
if (typeof process.stdin.unref === "function") process.stdin.unref();
}The caller already gates on !!process.stdin.isTTY, so this branch never executes. Remove it or add a comment explaining the defensive intent.
Minor: step number renumbering
The diff renumbers inline comments from 6–12 to 7–13 because the new gate was inserted. This is pure noise in the diff and makes git blame less useful. Consider dropping the sequence numbers or using labels.
What looks good
validateConfigDotpathis clean — prototype-pollution, empty segments, and numeric index guards all make sense- The troubleshooting doc entry follows one-sentence-per-line and covers all three modes (interactive, non-interactive, env var)
- The
--config-accept-new-pathflag name is discoverable and the error message tells you exactly what to do
- Extract classifyNewKeyGate as a pure helper so the gate is testable without mocking process.exit, stdin, or env. Add unit tests covering flag override, env override, env-not-exactly-"1", non-TTY refusal, NEMOCLAW_NON_INTERACTIVE=1 refusal, TTY prompt, and the precedence of overrides over non-interactive. - Drop the unreachable !process.stdin.isTTY branch in confirmYesNo (the caller already gates on isTTY). - Drop numeric prefixes from the inline step comments in configSet so future inserts no longer churn the surrounding lines and noise up git blame. - Annotate the lone ESM import to explain why it sits beside the file's existing CJS requires. Addresses review feedback on #2420. Signed-off-by: Tinson Lai <tinsonl@nvidia.com>
…nset-keys Signed-off-by: Tinson Lai <tinsonl@nvidia.com>
Contributor
There was a problem hiding this comment.
Actionable comments posted: 1
🤖 Prompt for all review comments with AI agents
Verify each finding against the current code and only fix it if needed.
Inline comments:
In `@src/lib/sandbox-config.ts`:
- Around line 421-451: The current "new key" gate treats oldValue === undefined
as a safe new path but misses the case where traversal failed because an
existing ancestor is a non-object scalar; setDotpath then clobbers that scalar.
Change the check around classifyNewKeyGate/oldValue to verify that every
existing ancestor along opts.key in config is an object (e.g., walk the dotpath
parents and ensure any present segment is typeof "object" and not null); only
treat it as a new path when oldValue === undefined AND all existing ancestors
are objects. If any existing ancestor is a non-object, refuse the write (use the
same error/exit flow instead of prompting) and include target.agentName and
opts.key in the message so the caller knows it would clobber a scalar; leave the
final setDotpath(config, opts.key, parsedValue) unchanged once the safety check
passes.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Enterprise
Run ID: 1ea52e0a-3891-4228-9f10-b17c7892dd53
📒 Files selected for processing (4)
.agents/skills/nemoclaw-user-reference/references/troubleshooting.mddocs/reference/troubleshooting.mdsrc/lib/sandbox-config.tstest/config-set.test.ts
✅ Files skipped from review due to trivial changes (2)
- .agents/skills/nemoclaw-user-reference/references/troubleshooting.md
- docs/reference/troubleshooting.md
…cestor setDotpath silently materialises plain objects at every intermediate segment, so a path like `provider.api` written against a config where `provider` is a string (or array, or null) used to overwrite the entire scalar with an empty object. The new gate also could not tell that case apart from a genuine new key. Add findClobberingAncestor, which walks the dotpath parents and reports the first existing segment that is not a config object so configSet can refuse the write with a descriptive message instead of silently clobbering. Numeric-segment refusal moves into the same helper now that both checks describe runtime safety of the same write — validateConfig- Dotpath stays purely syntactic (empty input, empty segments, prototype- pollution guards). Also switch the lone readline import to require so the file's module style stays consistent. Addresses review feedback on #2420. Signed-off-by: Tinson Lai <tinsonl@nvidia.com>
Contributor
There was a problem hiding this comment.
🧹 Nitpick comments (1)
src/lib/sandbox-config.ts (1)
750-762: Consider handling readline close event for robustness.The current implementation doesn't resolve the promise if the user presses Ctrl+C or stdin closes unexpectedly (the
closeevent fires without thequestioncallback). In practice, Ctrl+C triggers SIGINT which terminates the process, so this is rarely a problem for CLI usage.For robustness, you could handle the
closeevent:♻️ Optional: Handle readline close event
function confirmYesNo(prompt: string): Promise<boolean> { - return new Promise((resolve) => { + return new Promise((resolve, _reject) => { const rl = readline.createInterface({ input: process.stdin, output: process.stderr }); + let answered = false; + rl.on("close", () => { + if (!answered) resolve(false); + }); rl.question(prompt, (answer: string) => { + answered = true; rl.close(); resolve(/^y(es)?$/i.test(answer.trim())); }); }); }🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@src/lib/sandbox-config.ts` around lines 750 - 762, The confirmYesNo function can hang if stdin closes or the user triggers a readline 'close' without the 'question' callback firing; update confirmYesNo to attach an rl.on('close', ...) handler that resolves the promise (e.g., resolve(false)) and ensure you guard against double-resolution (use a local settled flag or remove listeners) and still close/cleanup the readline; reference the existing confirmYesNo function and the readline interface creation when adding this handler.
🤖 Prompt for all review comments with AI agents
Verify each finding against the current code and only fix it if needed.
Nitpick comments:
In `@src/lib/sandbox-config.ts`:
- Around line 750-762: The confirmYesNo function can hang if stdin closes or the
user triggers a readline 'close' without the 'question' callback firing; update
confirmYesNo to attach an rl.on('close', ...) handler that resolves the promise
(e.g., resolve(false)) and ensure you guard against double-resolution (use a
local settled flag or remove listeners) and still close/cleanup the readline;
reference the existing confirmYesNo function and the readline interface creation
when adding this handler.
ℹ️ Review info
⚙️ Run configuration
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Enterprise
Run ID: 2869a0f2-7c56-4584-b601-ce0231d4bfac
📒 Files selected for processing (2)
src/lib/sandbox-config.tstest/config-set.test.ts
🚧 Files skipped from review as they are similar to previous changes (1)
- test/config-set.test.ts
ericksoa
approved these changes
Apr 26, 2026
Contributor
ericksoa
left a comment
ericksoa
left a comment
There was a problem hiding this comment.
LGTM. Clean separation of pure helpers, strong test coverage, and security-critical guards (prototype-pollution, SSRF, gateway.*) preserved ahead of the new gate. Override path is discoverable for non-interactive use.
13 tasks
miyoungc
added a commit
that referenced
this pull request
Apr 28, 2026
## Summary Refreshes user-facing docs for the last 24 hours of merged NemoClaw history and bumps the docs metadata to 0.0.29, the next version after v0.0.28. The updates are limited to behavior supported by merged PR descriptions and diffs. ## Changes - `docs/reference/commands.md`: documented `nemoclaw <name> policy-add --from-file` and `--from-dir`, including custom preset review guidance, from #2077 / commit `7720b175`. - `docs/deployment/deploy-to-remote-gpu.md`: clarified that non-loopback `CHAT_UI_URL` disables OpenClaw device pairing for remote browser-only deployments, from #2449 / commit `f5ee8a4d`. - `docs/inference/inference-options.md`: documented provider-aware credential retry validation and the NVIDIA-only `nvapi-` prefix check, from #2389 / commit `6f7f0c6d`. - `docs/inference/switch-inference-providers.md`: documented `NEMOCLAW_INFERENCE_INPUTS` for text/image-capable model metadata baked into `openclaw.json`, from #2441 / commit `f4391892`. - `docs/reference/troubleshooting.md`: added the Git certificate verification entry for proxy CA propagation through `GIT_SSL_CAINFO`, `GIT_SSL_CAPATH`, `CURL_CA_BUNDLE`, and `REQUESTS_CA_BUNDLE`, from #2345 / commit `fa0dc1ab`. - `docs/versions1.json` and `docs/project.json`: promoted docs version `0.0.29`; `docs/versions1.json` omits unpublished `0.0.26`, `0.0.27`, and `0.0.28` entries. - `.agents/skills/nemoclaw-user-*`: regenerated derived user skill references from the updated docs. - Reviewed with no extra doc changes: #2575 / `d392ec07`, #2565 / `a3231049`, #1965 / `db1ef3ca`, #1990 / `db665834`, #2495 / `7da86fa3`, #2496 / `3192f4f4`, #2490 / `8c209058`, #2487 / `1f615e2f`, #2483 / `5653d33a`, #2482 / `31c782c0`, #2464 / `23bb5703`, #2472 / `a54f9a34`, and #2437 / `6bc860d7`. - Skipped per docs policy: #2420 / `7b76df6b` touched the experimental sandbox config path listed in `docs/.docs-skip`; #2466 / `cc15689c` touched a skipped term and CI-only sandbox image files. ## Type of Change - [ ] Code change (feature, bug fix, or refactor) - [ ] Code change with doc updates - [ ] Doc only (prose changes, no code sample modifications) - [x] Doc only (includes code sample changes) ## Verification <!-- Check each item you ran and confirmed. Leave unchecked items you skipped. --> - [x] `npx prek run --all-files` passes - [ ] `npm test` passes — failed locally in installer-integration tests and one onboard helper timeout; the doc-scoped hook test projects passed under `prek`. - [ ] Tests added or updated for new or changed behavior - [x] No secrets, API keys, or credentials committed - [x] Docs updated for user-facing behavior changes - [ ] `make docs` builds without warnings (doc changes only) — build succeeded, but local Sphinx emitted the existing version-switcher file read message. - [x] Doc pages follow the [style guide](https://github.com/NVIDIA/NemoClaw/blob/main/docs/CONTRIBUTING.md) (doc changes only) - [ ] New doc pages include SPDX header and frontmatter (new pages only) ## AI Disclosure <!-- If an AI agent authored or co-authored this PR, check the box and name the tool. Remove this section for fully human-authored PRs. --> - [x] AI-assisted — tool: Codex --- <!-- DCO sign-off required by CI. Run: git config user.name && git config user.email --> Signed-off-by: Miyoung Choi <miyoungc@nvidia.com> <!-- This is an auto-generated comment: release notes by coderabbit.ai --> ## Summary by CodeRabbit * **New Features** * Support for custom YAML presets in policy configuration via --from-file and --from-dir. * New build-time inference input option to declare accepted modalities (text or text,image). * **Improvements** * Credential validation now offers interactive recovery: re-enter key, retry, choose another provider, or exit. * Clarified provider-specific API key prefix handling (nvapi- only applies to NVIDIA keys). * **Documentation** * TLS certificate troubleshooting for inspected networks. * Clarified remote dashboard security/device-pairing behavior; command docs updated; docs version bumped. <!-- end of auto-generated comment: release notes by coderabbit.ai --> --------- Signed-off-by: Miyoung Choi <miyoungc@nvidia.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
The #1973 fix added a path validator that walked the currently-loaded config; #2400 reported that this rejected schema-valid first-time writes (the documented provider.compatible-endpoint.timeoutSeconds Ollama workaround) with the same error as a typo. An earlier attempt at a top-level allow-list improved discoverability but still couldn't catch deep typos, and would have had to be hand-maintained against OpenClaw's evolving schema.
This switches config set to a mode-based gate. Modifying an existing key still goes straight through. First-time writes prompt the user when stdin is a TTY, refuse with a discoverable opt-in when not, and accept an explicit override flag/env for scripts — so silent stub-writes stay impossible without coupling to a schema we don't own.
Related Issue
Fixes #2400
Changes
Type of Change
Verification
npx prek run --all-filespassesnpm testpassesmake docsbuilds without warnings (doc changes only)AI Disclosure
Signed-off-by: Tinson Lai tinsonl@nvidia.com
Summary by CodeRabbit