Tencent Cloud COS

• Outcome: Review
• Latest audit: May 29, 2026 · 9:27 PM UTC
• Version: 1.0.0

Overview

This is a coherent Tencent COS management skill, but it needs Review because it persistently stores powerful cloud credentials in plaintext and documents destructive deletion without enough safeguards.

Install only if you are comfortable granting this skill access to the target Tencent COS bucket. Use a dedicated least-privilege CAM key, avoid production-wide or delete-capable credentials unless needed, and prefer session-only environment variables or a secret manager over the default setup. Confirm every delete or recursive delete explicitly, list the target prefix first, and rotate any keys already written to shell startup files, mcporter.json, or coscmd config.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands
• Privilege Escalation: Excessive Permissions, Sudo/Root Execution, Credential Access
• Supply Chain: Unpinned Dependencies, External Script Fetching, Obfuscated Code
• Excessive Agency: Unrestricted Tool Access, Autonomous Decision Making, Scope Creep

Findings (6)

Missing User Warnings

Medium

• Confidence: 97% confidence
• Finding: The skill explicitly asks the user for highly sensitive Tencent Cloud credentials and instructs persisting them via setup.sh into shell startup files, but it does not clearly warn that the secrets will be stored on disk and remain available to future sessions. This creates a real risk of credential over-collection, accidental disclosure, or long-lived compromise if the local environment, shell history, or config files are exposed.

Missing User Warnings

Medium

• Confidence: 80% confidence
• Finding: The documentation exposes object deletion capability without any confirmation, warning, or guidance about irreversible data loss. In an agent skill context, this increases the chance that a user request is interpreted too broadly and destructive actions are taken without adequate safeguards.

Missing User Warnings

High

• Confidence: 95% confidence
• Finding: The skill documents recursive forced deletion commands using coscmd delete -r -f without any high-risk warning, confirmation requirement, or scope-limiting guidance. In an agent-operated environment, this can enable rapid mass deletion of cloud data from a single prompt or mistake, making the context materially more dangerous than ordinary CLI reference text.

Missing User Warnings

High

• Confidence: 91% confidence
• Finding: The documentation includes `coscmd delete -r <cosdir> -f`, which force-deletes directories recursively, but provides no prominent warning about irreversible bulk data loss. In an agent skill context, operators may treat the reference as safe-to-use guidance, increasing the chance of accidental mass deletion of cloud data.

Missing User Warnings

High

• Confidence: 97% confidence
• Finding: The setup script appends long-lived Tencent COS credentials directly into the user's shell startup file (~/.bashrc or ~/.zshrc) and exports them into the current session without any interactive warning, confirmation, or discussion of security implications. This creates persistent plaintext secret storage in a broadly accessed location that may be backed up, synced, inspected by other local processes, or accidentally exposed through dotfile sharing, making credential compromise materially more likely.

Missing User Warnings

High

• Confidence: 99% confidence
• Finding: The script writes SecretId and SecretKey into ~/.mcporter/mcporter.json as command-line arguments for cos-mcp, silently creating another plaintext copy of sensitive cloud credentials. This is especially risky because secrets embedded in config files and process arguments can be exposed via local file reads, backups, logs, process inspection, or accidental commits, expanding the attack surface beyond the user's shell environment.

VirusTotal

66/66 vendors flagged this skill as clean.

Static analysis

No suspicious patterns detected.