T09 · Insecure Skill Coding Practices
- Location
scripts/校验数据源配置.py:56- Finding
Access Token Exposure Through Command-Line Arguments and URL Query Parameters
- Content
View full analysis
- Remediation
View remediation
Security audit
Security checks for vulnerabilities and agentic risk
The skill is mostly purpose-aligned, but its scripts can expose access tokens and generated secrets despite documentation that downplays or contradicts those risks.
Install only if you are comfortable with an SDK integration assistant reading your project and proposing edits. Do not pass production tokens or secret_key values through command-line arguments or generated plan files unless you understand they may be retained in shell history, process metadata, /tmp files, source code, logs, backups, or commits. Review every diff, remove plaintext secrets where possible, gate SDK startup behind required privacy consent, and rotate any real token or secret already processed through these scripts.
scripts/校验数据源配置.py:56Access Token Exposure Through Command-Line Arguments and URL Query Parameters
scripts/generate_init_patch.py:707Secret Key Persisted in Generated Plan Files Despite Non-Persistence Claim
A mismatch between the declared purpose and actual behavior is a security concern because users may consent to an SDK integration helper while the skill performs more generic source processing behaviors. That gap undermines informed consent and can mask capabilities useful for wider codebase inspection or transformation beyond the advertised DataNexus-specific workflow.
Referenced artifact was not completely inspected
Step 7 自动校验 → scripts/validate_integration.py
Referenced artifact was not completely inspected
Step 7 自动校验 → scripts/validate_integration.py
Referenced artifact was not completely inspected
Step 7 自动校验 → scripts/validate_integration.py
The instructions direct developers to enable personal information collection and elsewhere request sensitive permissions, yet the guide lacks any user-facing disclosure, consent, retention, or compliance guidance. In the context of an SDK integration assistant, this is especially dangerous because developers may copy the example directly into production and begin collecting personal/device data by default, potentially violating privacy laws and app-store policies.
The skill describes capabilities that read projects, modify code, and invoke scripts, but it does not declare an explicit tool or permission scope. In an agent environment, that ambiguity can allow broader-than-expected file and network access, reducing containment and making accidental or unsafe actions more likely when handling user projects and credentials.
The skill is written entirely as a Chinese-language experience and includes a Chinese-only invocation/interaction design, but nowhere offers the user a language choice or states that the skill is restricted to Chinese for a documented regional reason. Under the language/locale policy rule, forcing a single language without opt-in is a natural-language policy concern.
The automated integration flow asks for sensitive values like secret_key and also authorizes code modification, but the skill text does not prominently warn about credential sensitivity, storage/echo risks, or the consequences of writing patches into the project. In practice this can lead users to disclose secrets in chat and approve repository changes without understanding the exposure or rollback implications.
The FAQ explicitly states that GDTAction.start() initiates personal/device information collection and presents it as a required step, but provides no warning about consent, disclosure, or privacy-law prerequisites. In an SDK integration assistant, this omission is risky because developers may copy the guidance directly and enable collection before obtaining valid user authorization, leading to unlawful or non-compliant data collection.
The FAQ discusses optional access to READ_PHONE_STATE for IMEI and OAID collection details, including certificate/version handling, but does not warn that these identifiers are sensitive and may require prominent disclosure, runtime permission handling, and jurisdiction-specific consent. In the context of an SDK onboarding guide, this increases the chance that developers implement identifier collection as a technical requirement without necessary privacy safeguards.
The document instructs developers to initialize telemetry collection, start personal-information collection, and report user behaviors, but it provides no privacy/compliance guidance, consent requirements, or warning that these APIs may transmit user or device data. In an SDK integration assistant context, this omission can lead developers to implement tracking without lawful notice, consent, minimization, or platform-specific compliance checks, creating real privacy and regulatory risk.
The guide labels READ_PHONE_STATE and WRITE_EXTERNAL_STORAGE as optional optimizations, but these permissions materially expand access to sensitive identifiers and device data. Describing them as benign attribution improvements can cause developers to over-request permissions without understanding the privacy and platform-policy implications, increasing unnecessary data exposure.
The documentation explicitly says GDTAction.start() is required and that it starts personal information collection, but it provides no privacy gating, consent prerequisite, or minimization guidance. In an SDK integration guide, presenting personal data collection as a mandatory startup step can lead integrators to collect device information immediately at app launch without a lawful basis or informed user consent, creating material privacy and compliance risk.
Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.
**A:** iOS 14 起需通过 ATT(App Tracking Transparency)框架请求用户授权:
1. 在 `Info.plist` 中添加 `NSUserTrackingUsageDescription` 描述
2. 在合适时机调用 `ATTrackingManager.requestTrackingAuthorization` 请求权限
3. SDK 会自动适配授权状态,未授权时使用其他标识方案
This markdown file describes multiple behavior-reporting interfaces such as app start, registration, login, purchase, sharing, and automatic event collection, all of which imply transmission of user activity data. Under the markdown-file criteria for SQP-2, the description should include a user-facing warning about privacy or data collection impact, but no such warning appears in this section.
The file explicitly documents +[GDTAction getCaid] as returning CAID information, which is a privacy-sensitive identifier-related capability. For markdown files, SQP-2 requires warnings when behavior could affect privacy, but the description provides no caution about consent, platform compliance, or sensitive-data handling.
The document promotes broad automatic telemetry collection and extensive custom event reporting, including user actions and commerce events, without any guidance on transparency, consent, data minimization, or compliance review. This creates a meaningful privacy vulnerability because implementers may over-collect behavioral data or sensitive business/user metadata and transmit it silently through the SDK. The skill context increases risk because it is an SDK onboarding guide whose primary purpose is to drive instrumentation adoption.
The guide explicitly instructs developers to capture and transmit full incoming URL values via the open_url field. URLs can contain sensitive query parameters, tokens, referral identifiers, deep-link payloads, or user-specific data, so sending them to analytics without minimization, consent, or warning creates a real privacy and data-exposure risk. In an SDK integration guide, this is more dangerous because developers may copy the example verbatim into production apps at scale.
The documentation explicitly instructs integrators to set persistent identifiers such as OpenId and UnionId and to report user behavior events, but it provides no privacy notice, consent prerequisite, data-minimization guidance, or transmission/security caveats. In an SDK integration assistant, this omission is materially risky because developers may copy these patterns directly into production and collect trackable personal data without proper legal basis, user disclosure, or secure handling.
The document instructs developers to embed a secret_key directly in client-side小游戏 SDK initialization code, which exposes a credential to every end user of the app. In a game/miniprogram client, secrets cannot be kept confidential; attackers can extract them from packaged code or runtime memory and then forge telemetry, impersonate the data source, or abuse backend trust tied to that key.
The document instructs developers to enable and use automatic event collection and to provide user identifiers such as openid and unionid, but it does not clearly warn that these identifiers and user behavior data will be transmitted to a third-party service. In an SDK integration guide, this omission can lead teams to deploy tracking without adequate privacy notice, consent checks, data minimization, or platform-specific compliance review, increasing legal and privacy risk.
This markdown file describes setting OpenId/UnionId and reporting user behavior events, which can affect user privacy and involve transmitting identifiable data. The description provides no warning or disclosure about data collection, user consent, or privacy impact.
The document promotes automatic event collection and upload, including lifecycle/page events and identifiers such as openid/unionid, but does not pair this with any explicit privacy notice, consent guidance, or data-minimization warning. In the context of an SDK integration guide, this omission can lead integrators to deploy tracking that violates platform privacy expectations or applicable data-protection requirements.
The guide instructs developers to configure the endpoint https://api.datanexus.qq.com but does not warn that SDK initialization and event reporting will transmit user behavior data to a remote service. In an SDK integration guide focused on analytics collection, omission of privacy, consent, and data handling implications can cause downstream integrators to deploy tracking without adequate disclosure, lawful basis, or compliance review.
The document instructs integrators how to collect and set persistent user and device identifiers such as OpenId, UnionId, business user IDs, click IDs, channel IDs, ANID, and CAID, but it does not include any privacy notice, consent, lawful-basis, retention, or platform-compliance guidance. In an SDK integration context, this omission can lead downstream developers to implement tracking without adequate user disclosure or consent, creating regulatory, policy, and privacy risk at scale.