ClawHub
Back to skill

Security audit

Scaffold

Security checks for vulnerabilities and agentic risk

Overview

This scaffold skill performs broad but disclosed project-creation actions, including local files, git setup, and a private GitHub push, with no evidence of hidden or destructive behavior.

Install only if you want a scaffold tool that can create local projects, store org defaults in your home directory, initialize git, create a private GitHub repository with your gh authentication, and push generated files. Review the confirmation details carefully, especially the target path, repository name, and any PRD or internal content that will be included.

A.I.G

Vulnerability Patterns
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs

SkillSpector

Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Rogue AgentSelf-Modification, Session Persistence
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (6)

Session Persistence

Medium
Category
Rogue Agent
Content
---
name: solo-scaffold
description: Generate complete project from PRD + stack template — directory structure, configs, CLAUDE.md, git repo, and GitHub push. Use when user says "scaffold project", "create new project", "start new app", "bootstrap project", or "set up from PRD". Uses SoloGraph for patterns and Context7 for latest versions. Do NOT use for planning features (use /plan) or PRD generation (use /validate).
license: MIT
metadata:
  author: fortunto2
Confidence
60% confidence
Finding
Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.

Missing User Warnings

Medium
Confidence
96% confidence
Finding
The skill performs high-impact side effects—creating local files, initializing git, creating a remote GitHub repository, and pushing code—but the top-level description does not clearly warn users about those actions. That creates a consent/transparency problem: a user invoking a seemingly generic scaffold command may not realize it will publish artifacts remotely.

Missing User Warnings

Medium
Confidence
95% confidence
Finding
The skill writes a persistent configuration file in the user's home directory (`~/.solo-factory/defaults.yaml`) without prominent up-front disclosure. Persistent writes outside the project directory can surprise users, alter future behavior, and store potentially sensitive organizational metadata without clear consent.

Skill Enumeration

Medium
Category
Agent Snooping
Content
**CLI-First Testing:** generate a `cli/` directory with a stub that imports core business logic from `lib/` (or equivalent). The CLI should run the main pipeline deterministically without requiring LLM, network, or UI. This enables `make integration` for pipeline verification. See `dev-principles.md` → "CLI-First Testing".

   ### `.claude/skills/dev/SKILL.md` — product dev workflow skill

   Generate a skill that teaches Claude how to work with THIS specific project. Structure:
Confidence
80% confidence
Finding
Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Skill Enumeration

Medium
Category
Agent Snooping
Content
10. **Generate CLAUDE.md** for the new project:
   - Project overview (problem/solution from PRD)
   - Tech stack (packages + versions from Context7)
   - **Skills section:** list available `.claude/skills/` with descriptions
   - Directory structure
   - Common commands (reference `make help`)
   - SGR / Domain-First section
Confidence
85% confidence
Finding
Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Session Persistence

Medium
Category
Rogue Agent
Content
Generated by /scaffold"
    ```

15. **Create GitHub private repo + push:**
    ```bash
    cd <projects_dir>/<name>
    gh repo create <name> --private --source=. --push
Confidence
83% confidence
Finding
Automatically creating a GitHub repository and pushing generated code causes session output to persist to a third-party remote service. In the context of a scaffold skill, this is especially sensitive because generated files may include PRD content, internal architecture notes, defaults-derived org identifiers, or other material the user did not intend to publish yet.

Static analysis

No suspicious patterns detected.