ClawHub
Back to skill

Security audit

Dynamic Project Soul Snapshot | 动态项目灵魂快照

Security checks for vulnerabilities and agentic risk

Overview

This skill is not malicious, but it can persist recent project and chat context into a shareable AI_CONTEXT.md file, so users should review it carefully before installing or sharing the output.

Install only if you intentionally want a project-local AI_CONTEXT.md snapshot. Invoke it explicitly by name, inspect the generated file before committing or sharing, run a secrets scan, and consider adding AI_CONTEXT.md to .gitignore for private, customer-sensitive, or credential-heavy work.

SkillSpector

Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • 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
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (3)

Vague Triggers

Medium
Confidence
86% confidence
Finding
The skill registers on broad everyday phrases like 'save my progress' and 'I'm done for today,' which can cause unintended activation. In this skill's context, accidental activation is riskier because it triggers repository scanning, git history collection, conversation mining, and persistence of the result into a shareable project file, increasing chances of sensitive context being captured without explicit user intent.

Ssd 3

Medium
Confidence
96% confidence
Finding
The skill explicitly instructs the agent to mine recent conversation history and then serialize extracted content into `AI_CONTEXT.md` for reuse. That creates a natural-language exfiltration and persistence channel for secrets, credentials, internal plans, and other sensitive user-provided data that may have appeared in chat, especially since the file is intended to be committed, shared, or pasted into future AI sessions.

Ssd 3

Medium
Confidence
95% confidence
Finding
The bootstrap block is specifically designed for copy-paste into a new AI session, making cross-session transmission of accumulated context a core feature. In context, that is dangerous because the same file also aggregates git diffs, decisions, traps, and potentially conversation-derived content, so sensitive internal information can be propagated to new tools, users, or external AI providers with minimal friction.

Static analysis

No suspicious patterns detected.