惯性聚合 高效追踪和阅读你感兴趣的博客、新闻、科技资讯
阅读原文 在惯性聚合中打开

推荐订阅源

F
Full Disclosure
Recorded Future
Recorded Future
T
Tenable Blog
S
Securelist
C
CERT Recently Published Vulnerability Notes
T
Threatpost
S
Schneier on Security
A
Arctic Wolf
The Hacker News
The Hacker News
C
CXSECURITY Database RSS Feed - CXSecurity.com
Know Your Adversary
Know Your Adversary
P
Privacy International News Feed
Threat Intelligence Blog | Flashpoint
Threat Intelligence Blog | Flashpoint
The Register - Security
The Register - Security
Cisco Talos Blog
Cisco Talos Blog
AWS News Blog
AWS News Blog
K
Kaspersky official blog
T
True Tiger Recordings
T
Threat Research - Cisco Blogs
V
Vulnerabilities – Threatpost
P
Palo Alto Networks Blog
T
The Exploit Database - CXSecurity.com
小众软件
小众软件
B
Blog
Cyber Security Advisories - MS-ISAC
Cyber Security Advisories - MS-ISAC
Microsoft Azure Blog
Microsoft Azure Blog
Cyberwarzone
Cyberwarzone
C
Cybersecurity and Infrastructure Security Agency CISA
T
Tor Project blog
Spread Privacy
Spread Privacy
Malwarebytes
Malwarebytes
P
Proofpoint News Feed
F
Fox-IT International blog
F
Fortinet All Blogs
P
Privacy & Cybersecurity Law Blog
G
GRAHAM CLULEY
量子位
Latest news
Latest news
OSCHINA 社区最新新闻
OSCHINA 社区最新新闻
博客园 - 叶小钗
Project Zero
Project Zero
T
Tailwind CSS Blog
N
Netflix TechBlog - Medium
Martin Fowler
Martin Fowler
IntelliJ IDEA : IntelliJ IDEA – the Leading IDE for Professional Development in Java and Kotlin | The JetBrains Blog
IntelliJ IDEA : IntelliJ IDEA – the Leading IDE for Professional Development in Java and Kotlin | The JetBrains Blog
I
Intezer
博客园_首页
腾讯CDC
H
Hackread – Cybersecurity News, Data Breaches, AI and More
D
Darknet – Hacking Tools, Hacker News & Cyber Security

Recent Commits to openclaw:main

fix(qa): kill timed out Matrix CLI runs · openclaw/openclaw@b8fc2f6 fix(dev): bound discord smoke waits · openclaw/openclaw@545ad7f chore(pixverse): publish as external plugin · openclaw/openclaw@5366209 feat(pixverse): add api region selection · openclaw/openclaw@b3083de feat(pixverse): add video generation provider · openclaw/openclaw@c183705 fix(qa): tolerate fast Matrix tool replies · openclaw/openclaw@a46e839 refactor: remove channel turn runtime aliases · openclaw/openclaw@6c37402 fix(test): bound qa otel receiver bodies · openclaw/openclaw@83ab0ba fix(lint): clean manifest registry installed checks · openclaw/openclaw@fd648ed fix(scripts): resolve npm package candidates through npm runner · openclaw/openclaw@7a7d9de fix(qa): hide Matrix tool progress marker in workspace · openclaw/openclaw@42f3550 fix(usage): forward cached token usage in chat completions (#82062) · openclaw/openclaw@12e5876 test(codex): align provider claim expectation · openclaw/openclaw@42387af perf(gateway): cache stable plugin index fingerprints · openclaw/openclaw@2babe03 fix(test): scan kitchen rpc readiness logs incrementally · openclaw/openclaw@1d4537a fix(test): fail startup bench on bad samples · openclaw/openclaw@8c6da93 fix(onboard): preserve agents.list and bindings on rerun · openclaw/openclaw@bbdff39 fix: send bare direct Anthropic model ids · openclaw/openclaw@aa0a290 fix(deepinfra): load all DeepInfra models when user wants to browse t… fix(package): honor dist package exclusions in inventory · openclaw/openclaw@296fbde fix(qa): harden Matrix tool progress scenario · openclaw/openclaw@32b3fb6 fix(package): match npm globstar exclusions · openclaw/openclaw@f4bcd61 fix: preserve channel runResolved mock compatibility · openclaw/openclaw@c89298f fix(test): bound config reload log polling · openclaw/openclaw@329dad2 fix(lint): preserve JSONL parse cause · openclaw/openclaw@d6949d5 test(e2e): preserve macos smoke entrypoint path · openclaw/openclaw@5eba765 fix(test): await mcp timeout cleanup · openclaw/openclaw@109ba23 fix(test): harden mcp channel ws timeout · openclaw/openclaw@a4a75a8 fix(test): harden gateway network ws timeout · openclaw/openclaw@e50b20f fix(agents/harness): validate forced plugin harness support before pi… · openclaw/openclaw@730ac1a fix(test): bound codex media path log polling · openclaw/openclaw@40a2600 test: align extension inbound context assertions · openclaw/openclaw@98c0ad8 fix(install): skip Homebrew until macOS packages need it · openclaw/openclaw@527b7c2 fix(ci): bound additional boundary checks · openclaw/openclaw@351aac9 fix(release): reject empty beta smoke runs · openclaw/openclaw@4dfc2cf fix(build): cap tsdown heap in containers · openclaw/openclaw@e8dde30 fix(docker): skip declarations in runtime packages · openclaw/openclaw@cc662ba fix(package): omit unpacked test helpers from inventory · openclaw/openclaw@6c42fea fix(crabbox): reinitialize invalid changed-gate git dirs · openclaw/openclaw@ecdc925 fix(lint): shard core lint checks · openclaw/openclaw@1ba4448 fix(lint): cap oxlint helper memory locally · openclaw/openclaw@8caa44f fix(lint): split source lint shards · openclaw/openclaw@158bc69 fix(crabbox): full-sync local sparse container runs · openclaw/openclaw@b3e3b1b fix(test): enable live cache script gates · openclaw/openclaw@72c6813 fix(test): reject unknown live media providers · openclaw/openclaw@51dd548 fix(cli): reject loose model and gateway numeric options · openclaw/openclaw@6b391ef refactor: centralize inbound supplemental context · openclaw/openclaw@1507a97 fix(test): reject empty gateway cpu runs · openclaw/openclaw@ad3d197 fix(test): fail empty plugin gauntlet runs · openclaw/openclaw@b460ee4 fix: load Claude CLI OAuth for PI auth profiles (#87167) · openclaw/openclaw@cc704ca fix(agents): ignore failed subagent placeholders fix(agents): report approval resolutions in bridge mode · openclaw/openclaw@88bbc5b fix(qa): keep fallback delivery on latest targets fix(agents): preserve bridge hook context · openclaw/openclaw@2c3190d fix(qa): close remaining mock qa e2e regressions · openclaw/openclaw@14198a1 fix(qa): isolate mock bridge hook state · openclaw/openclaw@35248be fix(qa): stabilize mock QA scenario contracts · openclaw/openclaw@81c1892 fix(agents): classify direct fallback targets by channel grammar · openclaw/openclaw@79f7b93 fix(qa): scope mock image prompts to latest turn · openclaw/openclaw@c2d059d fix(agents): suppress Write/Edit failed warning on response-timeout f… · openclaw/openclaw@7e702bb fix(cron): surface classified run failure causes · openclaw/openclaw@3104f36 fix(test): fail empty extension test requests · openclaw/openclaw@57b1c0b fix(cli): reject loose numeric options · openclaw/openclaw@c95d348 docs(providers/openai): clarify OpenAI Realtime Platform credits · openclaw/openclaw@717003a fix(codex): keep attempt watchdog for queued terminal turns · openclaw/openclaw@ca990f2 fix(qqbot): gate fallback approval buttons (#87154) · openclaw/openclaw@08a73db fix(test): fail explicit empty vitest runs · openclaw/openclaw@7615c31 ci: fall back from stale workflow dispatch refs · openclaw/openclaw@8d99037 fix(lint): serialize oxlint shards on constrained hosts · openclaw/openclaw@c93b7d8 fix: reject partial numeric CLI options fix(agents): avoid duplicate Claude CLI skill prompts · openclaw/openclaw@f4e20f8 test: harden e2e instance package fixture fix(codex): preserve raw reasoning source-reply guard · openclaw/openclaw@284098d test(codex): verify completion idle watch arms after non-assistant ra… · openclaw/openclaw@4d6bcf9 fix(codex): keep raw assistant release path intact · openclaw/openclaw@4314ead test(codex): mirror raw reasoning event order · openclaw/openclaw@e718d47 fix(codex): arm completion idle watch after rawResponseItem/completed… · openclaw/openclaw@a36c82b fix(node-host): restart stale node host on version mismatch · openclaw/openclaw@819fd9f fix(e2e): bound tool search gateway proof · openclaw/openclaw@761c802 fix(discord): harden requester checks for guild actions · openclaw/openclaw@9ed1b02 fix: let skills JSON output flush naturally · openclaw/openclaw@84b1123 fix: support plugin generated help targets · openclaw/openclaw@ec377dd fix: route generated help targets to subcommands · openclaw/openclaw@1de9848 fix: validate gateway call timeouts · openclaw/openclaw@f407e4e fix: mark plugin command groups in root help · openclaw/openclaw@482018e fix: route nested root help targets · openclaw/openclaw@2bbef6c fix: keep root help plugin descriptor loading quiet · openclaw/openclaw@ef2ebee fix: preserve root options in generated help · openclaw/openclaw@b31c9e9 fix: route root help targets to command help · openclaw/openclaw@5f6293a fix: normalize generated help self-help · openclaw/openclaw@9015d0c perf(secrets): propagate snapshots and eliminate esm side-effects in … test: stabilize main ci lanes · openclaw/openclaw@f327df8 perf(gateway): keep agent session working store active-only · openclaw/openclaw@54eca3f test(e2e): harden shell helper env assertions · openclaw/openclaw@e6937f9 fix(doctor): warn on unsupported active tool schemas · openclaw/openclaw@4976851 test: fix main ci expectations · openclaw/openclaw@3684696 fix(e2e): bound Telegram credential setup · openclaw/openclaw@9b28603 perf(gateway): borrow agent session lookup safely · openclaw/openclaw@3b9fa16 fix(crabbox): prefer Azure for Windows targets (#87186) · openclaw/openclaw@6afe3e8 test(e2e): isolate shell helper env · openclaw/openclaw@b28f9e0
fix(doctor): validate bundled MCP tool schemas · openclaw/openclaw@a02fe52
vincentkoc · 2026-05-27 · via Recent Commits to openclaw:main

@@ -7,12 +7,20 @@ import {

77

} from "../agents/model-catalog.js";

88

import { resolveDefaultModelForAgent } from "../agents/model-selection.js";

99

import { supportsModelTools } from "../agents/model-tool-support.js";

10+

import { createBundleMcpToolRuntime } from "../agents/pi-bundle-mcp-tools.js";

11+

import { applyFinalEffectiveToolPolicy } from "../agents/pi-embedded-runner/effective-tool-policy.js";

12+

import { shouldCreateBundleMcpRuntimeForAttempt } from "../agents/pi-embedded-runner/run/attempt-tool-construction-plan.js";

1013

import { createOpenClawCodingTools } from "../agents/pi-tools.js";

1114

import { normalizeAgentRuntimeTools } from "../agents/runtime-plan/tools.js";

1215

import { buildWorkspaceSkillStatus, type SkillStatusEntry } from "../agents/skills-status.js";

13-

import { inspectRuntimeToolInputSchemas } from "../agents/tool-schema-projection.js";

16+

import {

17+

inspectRuntimeToolInputSchemas,

18+

type RuntimeToolSchemaDiagnostic,

19+

} from "../agents/tool-schema-projection.js";

20+

import type { AnyAgentTool } from "../agents/tools/common.js";

1421

import { collectUnavailableAgentSkills } from "../commands/doctor-skills-core.js";

1522

import type { OpenClawConfig } from "../config/types.openclaw.js";

23+

import { formatErrorMessage } from "../infra/errors.js";

1624

import type { ProviderRuntimeModel } from "../plugins/provider-runtime-model.types.js";

1725

import { getPluginToolMeta } from "../plugins/tools.js";

1826

import type { HealthFinding } from "./health-checks.js";

@@ -56,6 +64,100 @@ function buildDoctorRuntimeModel(params: {

5664

} as ProviderRuntimeModel;

5765

}

586667+

function toolSchemaDiagnosticToFinding(params: {

68+

tools: readonly AnyAgentTool[];

69+

diagnostic: RuntimeToolSchemaDiagnostic;

70+

}): HealthFinding {

71+

const tool = params.tools[params.diagnostic.toolIndex];

72+

const pluginId = tool ? getPluginToolMeta(tool)?.pluginId : undefined;

73+

const owner = pluginId ? ` from plugin ${pluginId}` : "";

74+

const path =

75+

pluginId === "bundle-mcp"

76+

? "mcp.servers"

77+

: pluginId

78+

? `plugins.entries.${pluginId}`

79+

: `tools.${params.diagnostic.toolName}`;

80+

const fixHint =

81+

pluginId === "bundle-mcp"

82+

? "Disable or update the offending MCP server/tool so its parameters are a JSON object schema, then rerun doctor."

83+

: "Disable or update the offending plugin/tool so its parameters are a JSON object schema, then rerun doctor.";

84+

return {

85+

checkId: "core/doctor/runtime-tool-schemas",

86+

severity: "error",

87+

message: `Tool ${params.diagnostic.toolName}${owner} has an unsupported input schema for runtime projection.`,

88+

path,

89+

target: params.diagnostic.toolName,

90+

requirement: params.diagnostic.violations.join(", "),

91+

fixHint,

92+

};

93+

}

94+95+

function collectToolSchemaFindings(tools: readonly AnyAgentTool[]): HealthFinding[] {

96+

return inspectRuntimeToolInputSchemas(tools).map((diagnostic) =>

97+

toolSchemaDiagnosticToFinding({

98+

tools,

99+

diagnostic,

100+

}),

101+

);

102+

}

103+104+

async function collectBundleMcpRuntimeToolSchemaFindings(params: {

105+

cfg: OpenClawConfig;

106+

agentId: string;

107+

workspaceDir: string;

108+

modelRef: { provider: string; model: string };

109+

model: ProviderRuntimeModel;

110+

}): Promise<readonly HealthFinding[]> {

111+

if (

112+

!shouldCreateBundleMcpRuntimeForAttempt({

113+

toolsEnabled: true,

114+

})

115+

) {

116+

return [];

117+

}

118+119+

let bundleRuntime: Awaited<ReturnType<typeof createBundleMcpToolRuntime>> | undefined;

120+

try {

121+

bundleRuntime = await createBundleMcpToolRuntime({

122+

workspaceDir: params.workspaceDir,

123+

cfg: params.cfg,

124+

});

125+

const activeBundleTools = applyFinalEffectiveToolPolicy({

126+

bundledTools: bundleRuntime.tools,

127+

config: params.cfg,

128+

agentId: params.agentId,

129+

modelProvider: params.modelRef.provider,

130+

modelId: params.modelRef.model,

131+

warn: () => {},

132+

});

133+

const normalizedTools = normalizeAgentRuntimeTools({

134+

tools: activeBundleTools,

135+

provider: params.modelRef.provider,

136+

config: params.cfg,

137+

workspaceDir: params.workspaceDir,

138+

env: process.env,

139+

modelId: params.modelRef.model,

140+

modelApi: params.model.api,

141+

model: params.model,

142+

});

143+

return collectToolSchemaFindings(normalizedTools);

144+

} catch (error) {

145+

return [

146+

{

147+

checkId: "core/doctor/runtime-tool-schemas",

148+

severity: "error",

149+

message: "Configured MCP tool schema validation could not load the runtime tool set.",

150+

path: "mcp.servers",

151+

requirement: formatErrorMessage(error),

152+

fixHint:

153+

"Fix or disable the offending MCP server, then rerun doctor before relying on assistant tool startup.",

154+

},

155+

];

156+

} finally {

157+

await bundleRuntime?.dispose();

158+

}

159+

}

160+59161

export async function collectRuntimeToolSchemaFindings(

60162

cfg: OpenClawConfig,

61163

): Promise<readonly HealthFinding[]> {

@@ -97,19 +199,14 @@ export async function collectRuntimeToolSchemaFindings(

97199

modelApi: model.api,

98200

model,

99201

});

100-

return inspectRuntimeToolInputSchemas(normalizedTools).map((diagnostic) => {

101-

const tool = normalizedTools[diagnostic.toolIndex];

102-

const pluginId = tool ? getPluginToolMeta(tool)?.pluginId : undefined;

103-

const owner = pluginId ? ` from plugin ${pluginId}` : "";

104-

return {

105-

checkId: "core/doctor/runtime-tool-schemas",

106-

severity: "error",

107-

message: `Tool ${diagnostic.toolName}${owner} has an unsupported input schema for runtime projection.`,

108-

path: pluginId ? `plugins.entries.${pluginId}` : `tools.${diagnostic.toolName}`,

109-

target: diagnostic.toolName,

110-

requirement: diagnostic.violations.join(", "),

111-

fixHint:

112-

"Disable or update the offending plugin/tool so its parameters are a JSON object schema, then rerun doctor.",

113-

} satisfies HealthFinding;

114-

});

202+

return [

203+

...collectToolSchemaFindings(normalizedTools),

204+

...(await collectBundleMcpRuntimeToolSchemaFindings({

205+

cfg,

206+

agentId,

207+

workspaceDir,

208+

modelRef,

209+

model,

210+

})),

211+

];

115212

}