Back to skill

Security audit

社区运营小助手

Security checks for vulnerabilities and agentic risk

Overview

This looks like a real merchant operations skill, but it needs review because it stores a powerful merchant key insecurely and exposes broad data-query and write capabilities.

Install only if you trust the Koubei/OpenClaw service and are comfortable granting this skill a merchant key that can query business/user data and perform remote actions such as publishing, editing posts, sending bulk messages, changing labels, and increasing points. Use a least-privilege key if available, protect or rotate the key, and manually review every SQL query, file upload, message batch, points change, and post edit before allowing the agent to run it.

Vulnerability Patterns
  • Insecure Skill Coding PracticesFinds exploitable flaws such as hardcoded secrets or command injection
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
Findings (3)

T09 · Insecure Skill Coding Practices

Error
Location
scripts/openclaw.js:32
Finding

Merchant API Key Is Exposed Through Plaintext Storage, Process Arguments, and URL Query Parameters

Content
View full analysis
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Warning
Location
scripts/openclaw.js:98
Finding

Documented SQL Security Restrictions Are Not Enforced by the CLI

Content
View full analysis
console.error('错误:', e.message)); ``` ### Technical Analysis The Skill documentation establishes query restrictions, including: - Only exposed tables and fields may be queried. - Wildcard selection is prohibited. - The maximum result limit is 500. - Unsupported fields and `forum_id` must not be queried. - Required time-range predicates must match the user's request. - Post queries should default to approved posts. - Fields marked as non-searchable must not be used as filters. The CLI does not parse, validate, or constrain the SQL. It concatenates all remaining command-line arguments and forwards the resulting string directly to the remote query endpoint. This is a missing defense-in-depth control rather than proof that the remote database itself is directly injectable. Server-side validation may exist, but it is not visible in the audited project. If server-side enforcement is incomplete, a mistaken, compromised, or manipulated Agent can submit queries outside the documented data-access policy. ### Attack Path 1. An attacker supplies a crafted request or prompt that induces the Agent to construct an overbroad or prohibited SQL query. 2. The Agent invokes the `query` command with that SQL. 3. The CLI accepts the query without checking ...[truncated 880 chars]
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Warning
Location
scripts/openclaw.js:374
Finding

Sensitive Write and Bulk Operations Lack Programmatic Confirmation Controls

Content
View full analysis
0) { result.results.forEach(r => { if (r.status !== 'success') { console.log(` 用户 ID ${r.user_id}: 失败 (${r.message})`); } }); } return result; } ``` ### Technical Analysis The Skill text requires user confirmation for several write operations, but the executable interface does not enforce that requirement. Once invoked, the affected functions immediately submit their requests. There is no confirmation token, interactive confirmation prompt, dry-run mode, operation preview, recipient cap ...[truncated 1679 chars]
Remediation
View remediation
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (29)

Tp4

High
Category
MCP Tool Poisoning
Confidence
96% confidence
Finding

The skill advertises a relatively bounded merchant-operations purpose, but the instructions reveal materially broader capabilities including arbitrary SQL querying, access to table listings, bulk messaging, tag management, points adjustment, and management of "sockpuppet/马甲" accounts. This creates a trust-boundary problem: users or hosting platforms may authorize the skill based on a narrower description while unknowingly exposing sensitive data and high-impact write operations, enabling abuse, over-privileged use, and unintended data access.

Content

No source excerpt is available for this finding.

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 38)May include surrounding context.

md
node scripts/openclaw.js init --key {{key}}

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 61)May include surrounding context.

md
node scripts/openclaw.js init --key {{key}}

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 175)May include surrounding context.

md
node scripts/openclaw.js init --key {{key}}

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 184)May include surrounding context.

md
node scripts/openclaw.js init --key {{key}}

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 214)May include surrounding context.

md
node scripts/openclaw.js init --key {{key}}

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 251)May include surrounding context.

md
node scripts/openclaw.js init --key {{key}}

Description-Behavior Mismatch

High
Category
Not specified by scanner
Confidence
94% confidence
Finding

The documented CLI exposes materially broader capabilities than the skill description suggests, including arbitrary data-table querying, identity/account management, points adjustments, internal messaging, and user labeling. In an agent setting, this increases the chance that a user or prompt can drive high-impact actions on users and business data beyond the expected posting/operations scope, undermining least privilege and informed consent.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
Not specified by scanner
Confidence
97% confidence
Finding

The CLI exposes generic table enumeration and arbitrary SQL execution against the remote OpenClaw backend, which is substantially broader than the declared merchant-operations scope. If an agent or user can invoke this command, it can access or manipulate backend data outside intended workflows, creating a powerful data exfiltration and abuse primitive.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
Not specified by scanner
Confidence
93% confidence
Finding

The skill can alter user state by assigning labels, granting points, and creating/editing alias accounts, which are high-impact administrative actions not clearly disclosed in the manifest. In an agent skill, these functions can be abused for covert account manipulation, reputation gaming, or unauthorized incentive changes.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The trigger phrases are broad and map to ordinary conversational language such as configuration/help requests, increasing the chance the skill is invoked unintentionally in unrelated chats. In a skill that can store merchant keys, query backend data, and perform write operations, accidental invocation can lead to credential collection prompts, unintended disclosure, or execution of sensitive workflows under ambiguous user intent.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
86% confidence
Finding

Persistently binding a merchant key in a local config expands the trust boundary from per-operation authentication to ongoing credential storage on disk. That creates additional risk of credential leakage, reuse by unintended processes/users, and silent execution of privileged actions without fresh user awareness.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
88% confidence
Finding

The SQL guidance claims a default restriction to approved posts, but the example instead queries administrator user attributes, signaling that direct access to sensitive user/account tables is possible. In practice this weakens operator assumptions and may encourage prompts or agents to retrieve privileged user data under the guise of ordinary analytics.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
86% confidence
Finding

The direct upload commands can transmit arbitrary local files to a remote service, but the documentation provides no explicit warning that local file contents will leave the host. In an agent workflow, that omission raises the risk of accidental exfiltration of sensitive local data if a path is provided or inferred without clear user understanding.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
88% confidence
Finding

This markdown file includes a data export feature that allows downloading Excel/CSV files containing post, user, and conversion data, but it does not mention any caution about sensitive information, access control, or safe handling of exported data. For markdown files, SQP-2 applies when descriptions omit warnings about behaviors that could affect user data or privacy.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The README introduces an OpenClaw SQL/table-query capability that appears outside the stated Koubei Circle merchant-operations scope. This scope drift is dangerous because it normalizes broader data-access functionality, potentially enabling operators to query unrelated datasets and expanding the blast radius if the skill or its credentials are misused.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
86% confidence
Finding

The README instructs users to initialize and locally store an API key, but does not provide explicit credential-handling safeguards such as file-permission hardening, secret rotation, or prohibitions on sharing the config file. Because this skill requires merchant keys and performs operational actions, weak guidance around secret storage increases the risk of credential leakage and unauthorized use.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
88% confidence
Finding

The README explicitly demonstrates querying user-related fields such as user_id and nickname without any warning about personal data handling, authorization, or minimization. In a merchant-operations skill, this makes unsafe data access seem routine and can lead to privacy violations, overcollection, or improper downstream use of identifiable user information.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The API key is stored in plaintext in a local config.json, increasing exposure through local file disclosure, accidental commits, backups, or multi-user system access. Because this key authorizes merchant operations, compromise could enable unauthorized posting, messaging, and administrative actions.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

When a local file path is provided, the script reads the file and transmits its contents to the configured remote OpenClaw host. Although there is a log that an upload is happening, there is no warning about data leaving the local system or guidance to verify the target host before sending potentially sensitive media.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

This path reads a local video file and uploads it over HTTPS to the configured host. The console output notes an upload is occurring, but it does not clearly warn that local data is being sent off-machine or identify the destination, which is important for privacy-sensitive operations.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
89% confidence
Finding

The manifest lists posting, activity management, data monitoring, automatic reminders, and long-connection push, but does not mention direct outbound user messaging. sendStationMail performs batch站内信 delivery to arbitrary user IDs, which is a distinct engagement/moderation capability rather than an obvious implementation detail of posting automation.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
88% confidence
Finding

Bulk messaging is a high-impact action affecting many users, yet the CLI executes immediately without confirmation, dry-run, or recipient validation safeguards. Mistakes or abuse can quickly result in spam, phishing, harassment, or reputational damage at scale.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
86% confidence
Finding

This function updates an existing thread's title and/or message through a POST request, changing remote content. There is no user-facing confirmation, preview, or warning that the action will overwrite existing post content.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

Bulk point increases change user entitlements or rewards and are applied immediately to multiple users without any confirmation or anti-mistake safeguards. In this context, erroneous or malicious use can distort incentives, create financial/reputational impact, and undermine platform trust.

Content

No source excerpt is available for this finding.