chatgpt-app-builder
Pass
Audited by Gen Agent Trust Hub on May 13, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands to manage the development lifecycle, including project scaffolding (
npm create skybridge), dependency installation (npm install), and starting development servers. It also uses vendor-specific CLI tools for deployment (alpic deploy) and network tunneling (alpic tunnel) to facilitate testing with AI assistants. - [EXTERNAL_DOWNLOADS]: Orchestrates the retrieval of framework components and utility packages from well-known registries (npm) and vendor domains. These downloads are necessary for the skill's primary purpose of building and deploying apps within the Skybridge ecosystem.
- [PROMPT_INJECTION]: Identifies a surface for indirect prompt injection (Category 8) because the skill's workflow depends on reading and interpreting user-controlled specification files (
SPEC.md). - Ingestion points: Project requirements and architectural designs are parsed from
SPEC.mdto guide the agent's code generation. - Boundary markers: There are no explicit delimiters or specific instructions provided to the agent to treat the contents of
SPEC.mdas untrusted data. - Capability inventory: The skill possesses capabilities including file system modification, shell command execution, and establishing network tunnels.
- Sanitization: The skill lacks explicit validation or sanitization routines for the content of user-provided specification files before they are incorporated into the agent's context.
Audit Metadata