Back to skill

Security audit

guaikei-kuaishou-live-data-fetcher

Security checks across malware telemetry and agentic risk

Overview

The skill mostly matches its Kuaishou research purpose, but it merits Review because it automatically stores scraped results and tells users to print an API token.

Install only if you are comfortable sending Kuaishou research targets and your Guaikei API token to www.guaikei.com, and treat the generated logs directory as containing potentially sensitive scraped data. Do not print or share GUAIKEI_API_TOKEN; check only whether it is set, and rotate the token if it has already been exposed in logs or screenshots.

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

Warning
Location
SKILL.md:269
Finding
API Token Disclosure Through Troubleshooting Command## Vulnerability Details **File Location**: `SKILL.md:269-271` **Vulnerability Type**: Sensitive credential exposure through terminal output **Risk Level**: Medium **Vulnerable Code Snippet**: ```markdown **Q7. Command starts and exits immediately without returning data?** > This is usually because `GUAIKEI_API_TOKEN` failed validation (see Q1). Before running, use `echo $GUAIKEI_API_TOKEN` to confirm that the variable has been injected. ``` The original instruction at line 271 explicitly directs the user to run: ```bash echo $GUAIKEI_API_TOKEN ``` ### Technical Analysis `GUAIKEI_API_TOKEN` is an authentication credential used in the `TOKEN` HTTP header when communicating with `www.guaikei.com`. Printing the environment variable reveals its complete value in plaintext. Terminal output may be retained in CI/CD logs, terminal recordings, remote-session logs, shell transcripts, diagnostic bundles, screenshots, or screen-sharing sessions. The command is unnecessary for determining whether the variable is configured because its presence can be checked without exposing its value. The application code itself does not log the token, and sending it over HTTPS to the documented API provider is consistent with the Skill's declared authenticated data-retrieval functionality. The vulnerability is limited to the troubleshooting instruction. ### Attack Path 1. A user encounters a token-validation or startup failure. 2. The user follows the troubleshooting guidance in `SKILL.md`. 3. The user runs `echo $GUAIKEI_API_TOKEN`. 4. The complete token appears in terminal output. 5. The output is captured by a CI system, terminal recorder, remote support session, screenshot, screen share, or copied diagnostic transcript. 6. A party with access to that output obtains the token. 7. The party can submit authenticated requests to the Guaikei API using the exposed credential until it expires or is revoked. ### Impact Assessment ...[truncated 555 chars]
Remediation
## Remediation Suggestions Replace the credential-revealing command with a presence-only check: ```bash if [ -n "${GUAIKEI_API_TOKEN:-}" ]; then echo "GUAIKEI_API_TOKEN is configured" else echo "GUAIKEI_API_TOKEN is missing" fi ``` For debugging token formatting, report only non-sensitive metadata, such as length, and avoid printing any token characters: ```bash if [ -n "${GUAIKEI_API_TOKEN:-}" ]; then printf 'GUAIKEI_API_TOKEN is configured; length=%s\n' "${#GUAIKEI_API_TOKEN}" else echo "GUAIKEI_API_TOKEN is missing" fi ``` Additional hardening measures: - Explicitly warn users never to paste complete tokens into support tickets, chat messages, screenshots, or logs. - Redact authentication headers in all current and future debug logging. - Provide a documented token revocation and rotation process. - Advise affected users to rotate the token if they previously followed the vulnerable instruction in a logged or shared environment. - Use short-lived, narrowly scoped API tokens where supported.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (4)

Description-Behavior Mismatch

Medium
Confidence
89% confidence
Finding
The package manifest presents the skill as a Kuaishou analytics and research tool, but the exposed npm scripts include posting and commenting operations. This capability mismatch is security-relevant because users or orchestrators may grant the skill trust and permissions appropriate for read-only analysis while it also contains write actions that can perform account activity on external platforms.

Description-Behavior Mismatch

Medium
Confidence
90% confidence
Finding
The CLI writes fetched raw comment data to a local JSON file named from the target URL/video identifier without any consent prompt, minimization, or retention controls. User comments can contain personal or sensitive content, so silent persistence increases the risk of unintended collection, local exposure, and secondary reuse beyond the stated analytics purpose.

Missing User Warnings

Medium
Confidence
91% confidence
Finding
The README states that all task results are automatically saved to the logs/ directory, but it does not prominently warn users that scraped platform data, comments, URLs, and potentially sensitive research targets will be persisted locally. In a data-collection tool, silent persistence increases the risk of unintended retention, local exposure, and mishandling of collected content, especially on shared machines or CI/server environments.

Missing User Warnings

Medium
Confidence
92% confidence
Finding
The tool fetches remote comment content and automatically persists it to disk with no explicit notice or confirmation, which can surprise operators and create a local cache of potentially sensitive third-party data. In an analytics/research skill, this is more concerning because bulk collection and retention of user-generated content can exceed user expectations and increase privacy and compliance risk.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

Detected: suspicious.exposed_secret_literal

File appears to expose a hardcoded API secret or token.

Critical
Code
suspicious.exposed_secret_literal
Location
SKILL.md:15