Back to skill

Security audit

服务分发购物SKILL

Security checks for vulnerabilities and agentic risk

Overview

This shopping skill needs Review because it combines payment/scheduled-purchase authority with broad activation, weak credential handling, unsafe command/file patterns, no clear cancellation path, and a mock purchase path that reports success.

Only install after review and remediation. Require narrow explicit purchase intent, a final confirmation showing item, price, address, time, and total charge, a cancellation path for scheduled jobs, secure secret storage, fixed safe state paths, and a real provider-confirmed transaction flow or clear simulation labeling.

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 (4)

T09 · Insecure Skill Coding Practices

Error
Location
SKILL.md:42
Finding

Shell Command Injection Through Unquoted User-Controlled Shopping Intent

Content
View full analysis
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Error
Location
scripts/seap-cli.js:124
Finding

Arbitrary File Write Through Unvalidated Session Identifier

Content
View full analysis
Remediation
View remediation

other

Error
Location
scripts/seap-cli.js:57
Finding

Fabricated Purchase Success Misrepresents Financial Transactions

Content
View full analysis
item.skuId === skuId); if (!validSku) { return { success: false, orderId: "", message: `Invalid SKU: ${skuId}` }; } return mockBuyResponse; } ``` ### Technical Analysis The `aipay` command is described as completing a purchase, but it does not contact a merchant, payment provider, or cloud API. It does not read the configured authorization code, payment token, or delivery address. For any SKU contained in the fixed mock catalog, it returns a successful result with a random order identifier. The skill workflow instructs the agent to interpret the `success` field as proof of purchase and report completion to the user. Consequently, simulation output is presented as a real financial transaction without a test-mode warning or verifiable provider response. The response object is initialized once per process, so repeated calls in the same process would also reuse the same randomly generated order identifier. ### Attack Path 1. A user selects one of the fixed valid mock SKUs. 2. The agent invokes the documented `aipay` command. 3. `scheduleBuyIntention` checks only whether the SKU exists in local mock data. 4. The function returns the prebuilt `success: true` response without performing a transaction. 5. The agent relies on the success field and tells the user that the purchase completed. 6. The user may rely on a nonexistent order or perform a duplicate manual purchase after discovering the failure. ### Impact Assessment This issue compromises trans ...[truncated 404 chars]
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Warning
Location
seap.config.json:4
Finding

Plaintext Personal Delivery Address Committed in Configuration

Content
View full analysis
Remediation
View remediation
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • 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
Findings (12)

Vague Triggers

High
Category
Not specified by scanner
Confidence
98% confidence
Finding

The trigger condition activates on any message containing the single character “买”, which is overly broad for a skill that can search for products and ultimately place purchases. In this skill’s context, accidental activation is especially risky because it can begin a commerce flow and, combined with scheduled purchasing, could lead to unintended orders or user confusion around financial actions.

Content

No source excerpt is available for this finding.

Missing User Warnings

High
Category
Not specified by scanner
Confidence
96% confidence
Finding

The documentation describes scheduled purchases and automatic execution but does not prominently warn that real orders will be placed later using stored payment credentials and delivery data. For a skill that can spend funds without the user being present at execution time, this omission increases the chance of uninformed consent and unsafe use.

Content

No source excerpt is available for this finding.

Vague Triggers

High
Category
Not specified by scanner
Confidence
98% confidence
Finding

The skill auto-triggers whenever the message contains the single character "买", which is far too broad for a capability that can search products and initiate real purchases. This creates a high risk of accidental activation from casual conversation, quoted text, or ambiguous requests, especially because the skill supports unattended scheduled purchases.

Content

No source excerpt is available for this finding.

Missing User Warnings

High
Category
Not specified by scanner
Confidence
97% confidence
Finding

The skill documents storage of sensitive data including a payment token and delivery address, but does not present explicit warnings, handling constraints, or protections for those secrets. Because the token enables later unattended payment, insecure storage or casual handling could expose financial credentials and personal address information.

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 34)May include surrounding context.

md
依赖 `scripts/seap-cli.js` 执行相关命令。

Missing User Warnings

High
Category
Not specified by scanner
Confidence
97% confidence
Finding

The skill can execute actual purchases, including scheduled unattended purchases, yet the description does not clearly warn operators or users about this high-risk capability. In context, this omission is especially dangerous because the workflow combines broad triggering, persistent payment tokens, and deferred execution, which can enable unintended or unauthorized transactions.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
76% confidence
Finding

The README is written entirely in Chinese and all trigger and usage examples assume Chinese input, such as matching the character “买”. There is no statement that the skill is intentionally China/Chinese-specific or that users may choose another language, which can violate language/locale policy if the skill is expected to support broader audiences.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The README instructs users to place authCode and secretToken directly into a config file but does not warn that these are sensitive secrets that must be protected from disclosure, logging, version control, or insecure file permissions. Because the token is used for payment-related automation, compromise could expose account access or enable unauthorized purchases.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The manifest trigger phrase is ambiguous and insufficiently constrained, making it easier for the skill to activate in contexts that do not reflect a clear purchase authorization. In a shopping skill tied to payment operations, ambiguous activation materially increases the chance of unintended workflow entry and downstream unsafe actions.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

This code hardcodes Chinese-language output labels such as the heading, price, description, rating, store, and image fields. The file does not offer any language selection or explain that the skill is intentionally region-specific, which creates a locale-policy issue under the natural-language policy rules.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The CLI performs an unexpected side effect by writing search and purchase results to local disk, even though its stated purpose is shopping. In an agent/skill context, undocumented filesystem writes expand the capability surface and can leak data, overwrite files, or create persistence artifacts that downstream components did not expect.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
98% confidence
Finding

The sessionId is user-controlled and is directly interpolated into a filesystem path for writeFileSync without validation. An attacker can supply path traversal or absolute path values to write JSON files outside the intended directory, potentially overwriting application files or planting data in sensitive locations accessible to the process.

Content

No source excerpt is available for this finding.