ClawHub
Back to skill

Security audit

550W视频去字幕

Security checks for vulnerabilities and agentic risk

Overview

This skill coherently processes user-supplied media through the 550W service, with disclosed credentials, uploads, and remote URL probing, though users should be careful about which files and links they provide.

Install only if you trust the 550W service with the media and account credentials you provide. Do not pass sensitive local file paths or internal/private URLs unless you intend the host to read or fetch them, and be aware that some successful operations consume account credits.

SkillSpector

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
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (5)

Context-Inappropriate Capability

Medium
Confidence
93% confidence
Finding
The skill invokes an external ffprobe binary against user-supplied remote URLs, which expands capability from simple API orchestration into subprocess execution and outbound network access. This can enable SSRF-like behavior against internal or sensitive network targets reachable from the host, and it also increases attack surface through dependency on a local executable.

Context-Inappropriate Capability

Medium
Confidence
97% confidence
Finding
The CLI helper accepts arbitrary filePath input, resolves it, and reads any local file as a stream for upload. In an agent setting, this grants the skill filesystem read access beyond its stated remote media-processing purpose and can be abused to exfiltrate local secrets or sensitive user data.

Missing User Warnings

Medium
Confidence
95% confidence
Finding
The skill reads local files and prepares them for upload without any disclosure or consent prompt in code paths handling filePath. In an agent environment, silent local file access materially increases the chance of unintended exfiltration because users may believe they are only processing a named media item, not authorizing arbitrary filesystem reads.

Missing User Warnings

Medium
Confidence
90% confidence
Finding
When processing a remote video URL, the workflow probes that URL via ffprobe without any disclosure that the agent will make a direct outbound fetch from the host. This is dangerous because it can be used to contact attacker-controlled or internal endpoints and leaks network reachability from the execution environment.

Vague Triggers

Medium
Confidence
85% confidence
Finding
The manifest describes broad media-processing capabilities and exposes many actions without clearly constraining when the skill should trigger or what inputs are in scope. In agent ecosystems, vague capability descriptions can cause over-invocation on user media or links, increasing the chance of unintended handling of sensitive content, credential use, or abusive task submission.

Static analysis

Detected: suspicious.env_credential_access

Environment variable access combined with network send.

Critical
Code
suspicious.env_credential_access
Location
dist/550w-skill.cjs:104