ClawHub
Back to skill

Security audit

gingiris-opensource

Security checks across malware telemetry and agentic risk

Overview

This is a disclosed open-source marketing playbook, not a tool that installs code, reads private data, or runs actions on its own.

Installers should treat this as a marketing strategy guide. Review platform rules before using the DM, Reddit, Product Hunt, or pricing-page tactics, and require explicit user approval before posting publicly or messaging people from any real account.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • 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
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (4)

Vague Triggers

Medium
Confidence
94% confidence
Finding
The manifest-style trigger definitions include broad phrases such as "GitHub stars", "developer marketing", "open source growth", and especially "オープンソース", which are generic topic terms rather than narrowly scoped invocation phrases. Without exclusion conditions or clearer activation boundaries, the skill could be invoked during ordinary conversation about open source or GitHub rather than when the user wants this specific growth-playbook skill.

Natural-Language Policy Violations

Medium
Confidence
91% confidence
Finding
The instructions explicitly require the README and community channels to be "English-first," and later reinforce "English > multilingual > Chinese-only" as a default operating rule. This is a natural-language locale policy that privileges a specific language without presenting it as a user choice or clearly limiting it to a justified region-specific compliance context.

Unrestricted Tool Access

Medium
Category
Excessive Agency
Content
- No single country/region should exceed **20%** of your total stars.
- Global healthy split reference: China 19-21%, US 19-21%, rest scattered across Europe / Russia / Canada / Indonesia.
- Finer benchmark (from studying VSCode / Vue / AppFlowy the week before AFFiNE's launch): US 19-21%, China ~19-21%, FR+DE+IT+UK combined 10-15%; Russia and Brazil usually appear in the Top-10 countries. Week-1 stars matching this curve = healthy global cold start.
- Tools: **star-history.com** (trend chart) · **oss.cool / OSS Insight** (per-country breakdown, built by Chinese devs)
- **Launch sequence**: Week 1 → overseas-only outreach. Week 2 → domestic (Chinese) outreach. Reversing this breaks the distribution and kills Trending eligibility.
- AFFiNE real case: deliberately skipped WeChat Moments / Chinese community posts in week 1.
Confidence
85% confidence
Finding
Tools: *

Autonomous Decision Making

Medium
Category
Excessive Agency
Content
### PR article strategy

- Publish one authoritative article (official blog / Medium / dev.to) that fully explains the project's value.
- Purpose: give anyone who wants to spread your project a ready-made script. They will use it without asking.
- AFFiNE case: a German user independently posted to 40+ forums with no prior contact — only possible because a clear reference article existed.

### Competitor follower DMs
Confidence
75% confidence
Finding
without asking

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.