ClawHub
Back to skill

Security audit

strength-training-recording-summary

Security checks for vulnerabilities and agentic risk

Overview

This skill is a coherent workout transcript/audio summarizer with disclosed optional helpers and cloud transcription paths, though users should note the exercise-matching helper may fetch and cache a public dataset.

Installers should be comfortable with optional local shell tools and optional cloud transcription. Use an existing transcript or local Whisper for privacy-sensitive sessions, and pass a local exercises.json dataset if you do not want the matcher to fetch and cache the public GitHub dataset.

SkillSpector

Vulnerability Patterns
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • 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
Findings (4)

Lp3

Medium
Category
MCP Least Privilege
Confidence
87% confidence
Finding
The skill advertises a document-processing workflow but includes capabilities implying file access, shell execution, network access, and file writing without any declared permissions or clear user-facing constraints. This expands the attack surface because helper scripts, transcription backends, and remote dataset interactions could access local files or external services in ways the user did not explicitly authorize.

Tp4

High
Category
MCP Tool Poisoning
Confidence
94% confidence
Finding
The stated purpose is to summarize workout recordings or transcripts, but the analyzed behavior indicates additional remote downloading and fuzzy-matching logic while failing to perform the promised core summarization functions. Description-behavior mismatch is dangerous because users may trust the skill with sensitive recordings or transcripts under false assumptions, while hidden network activity can exfiltrate data, introduce supply-chain risk, or produce unexpected side effects.

Description-Behavior Mismatch

Medium
Confidence
94% confidence
Finding
The helper will download a JSON dataset from a remote GitHub URL and persist it under the user's home cache directory when no local dataset is provided. That creates undeclared network access and a supply-chain/trust boundary: remote content can change over time, affect outputs, and leave local artifacts, which is broader behavior than a recording/transcript summarizer would reasonably need.

Context-Inappropriate Capability

Medium
Confidence
91% confidence
Finding
The hardcoded DEFAULT_DATASET_URL introduces built-in outbound network capability to a third-party source for a skill whose declared purpose is summarizing workout recordings/transcripts. In this context, the mismatch increases risk because the skill can unexpectedly contact external infrastructure and consume unpinned data that may be modified or unavailable, affecting privacy, reliability, and trust.

Static analysis

No suspicious patterns detected.