Back to skill

Security audit

60s API 综合技能

Security checks for vulnerabilities and agentic risk

Overview

The skill is mostly a disclosed public-information API helper, but it needs review because it can send real passwords to an outside service.

Install only if you are comfortable with queries being sent to the 60s API service. Do not use the password-check endpoint with real passwords, API keys, tokens, or other secrets; prefer local password-strength checks instead. Treat translation, hash, QR, and OG URL features as external submissions of the provided text or URL.

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 (1)

T09 · Insecure Skill Coding Practices

Error
Location
SKILL.md:268
Finding

Remote Password Disclosure Through a URL Query Parameter

Content
View full analysis

Vulnerability Details

File Location: SKILL.md:268
Vulnerability Type: Plaintext sensitive-data transmission to a third-party API
Risk Level: High

Vulnerable Code

bash
curl -s "https://60s.viki.moe/v2/password/check?password=abc123"  # Password strength check

The endpoint is also documented in the API reference at SKILL.md:348:

markdown
| /v2/password/check | GET | password | Password strength check |

Technical Analysis

The Skill directs the Agent to send a password to the remote 60s.viki.moe service in a GET query parameter. Although HTTPS protects the request while it is in transit, it does not prevent the complete URL from being recorded by the remote server, reverse proxies, observability systems, diagnostic tooling, or other infrastructure with access to request metadata.

Invoking the example through a shell also risks exposing the password through command history, process inspection, execution traces, or Agent/tool logs. Sending a credential to an external service is unnecessary for password-strength analysis because policy, pattern, and entropy checks can be performed locally.

This behavior exceeds least privilege: the declared operation only requires evaluating password strength, but the implementation discloses the password to a third party.

Attack Path

  1. A user asks the Agent to assess the strength of an actual password.
  2. Following the Skill documentation, the Agent constructs a request to /v2/password/check.
  3. The plaintext password is inserted into the URL's password query parameter.
  4. The Agent invokes the request through curl and the execution tool.
  5. The URL may be retained in local command or tool logs and in remote API, proxy, monitoring, or diagnostic logs.
  6. An attacker or unauthorized operator with access to any retained log obtains the password.
  7. If the user reused the password, the exposed credential may be tested against oth ...[truncated 514 chars]
Remediation
View remediation

Remediation Suggestions

  1. Remove the remote password-strength endpoint and its example from the Skill instructions and API reference.
  2. Perform password-strength evaluation locally using length, character diversity, repetition, common-pattern, dictionary, and estimated-entropy checks.
  3. Explicitly instruct the Agent never to submit passwords, API keys, tokens, or other secrets to third-party services.
  4. Never place sensitive data in URL query parameters. If a remote operation is unavoidable, require explicit informed user consent and use a vetted service with a documented privacy policy and a request body, while recognizing that this still discloses the secret to that service.
  5. Prevent secret-bearing commands and request bodies from entering shell history, process listings, Agent traces, telemetry, and debug logs.
  6. Add a warning that users should not provide real credentials and that any sample must use an obviously synthetic value.
  7. Add automated linting or policy checks that reject Skill instructions containing credential-like parameters in URLs.
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 (7)

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The description states the skill triggers when users ask for various kinds of "中文信息", and the document repeatedly frames the skill as a Chinese-information skill without mentioning any user opt-in or language selection. This is a natural-language locale constraint that may violate language-choice policy because it imposes a specific language by default.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The activation section says the skill should activate whenever the user mentions broad keywords like "新闻", "天气", "人民币", "豆瓣", "翻译", or "颜色". These terms are common in ordinary conversation and the file does not provide exclusion conditions or negative examples, so the trigger scope is ambiguous and could cause unintended invocation.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
60% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · SKILL.md (reported line 172)May include surrounding context.

bash
# 汇率
curl -s "https://60s.viki.moe/v2/exchange-rate?from=USD&to=CNY"

# 农历(开发中,可能不稳定)
curl -s "https://60s.viki.moe/v2/lunar?date=2024-01-15"

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The skill explicitly instructs agents to use exec to run shell commands, even though its stated purpose is simple information retrieval from a web API. Allowing shell execution broadens the capability surface from HTTP fetches to arbitrary command execution, which can enable command injection or unsafe file/process interactions if any user-controlled values are interpolated into those commands.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
85% confidence
Finding

The generic instruction to send POST bodies via curl -X POST ... -d '{...}' encourages transmitting arbitrary user-provided content to an external service, and when combined with the exec guidance it raises both data-exposure and command-construction risks. This is more dangerous in this skill because several endpoints accept free-form text or URLs, which can include sensitive user input or be mishandled during shell assembly.

Content

Scanner excerpt · SKILL.md (reported line 377)May include surrounding context.

md
1. 所有接口免费,无需 API Key
2. 优先使用 `curl` 通过 `exec` 调用
3. POST 请求:`curl -X POST ... -H "Content-Type: application/json" -d '{...}'`
4. 图片/音频接口用 `-sL ... -o file` 保存
5. 展示结果精简到 TOP 5~10
6. 出错时重试一次,仍失败告知用户

Context-Inappropriate Capability

Low
Category
Not specified by scanner
Confidence
84% confidence
Finding

The OG metadata feature causes the agent to fetch arbitrary user-supplied URLs, expanding the skill from curated public-information lookups into general remote content retrieval. That creates SSRF-style risk, internal-network probing potential, and unexpected transmission of user-provided targets to a third-party service.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Low
Category
Not specified by scanner
Confidence
79% confidence
Finding

This file presents itself as a consolidated skill based on the 60s API and documents its own weather endpoints in detail. The note that '天气查询仍以 Open-Meteo 为主' introduces a conflicting implementation intent, implying the skill should rely on a different service than the one the rest of the document claims to provide.

Content

No source excerpt is available for this finding.