improve-codebase-architecture
Pass
Audited by Gen Agent Trust Hub on Jul 27, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests untrusted data from the codebase,
CONTEXT.md, and ADR files to inform its architectural review. - Ingestion points: Codebase files (via
Exploresubagent),CONTEXT.md, anddocs/adr/*.mdfiles. - Boundary markers: The instructions do not specify any delimiters or warnings to ignore instructions embedded within the codebase or documentation files.
- Capability inventory: The skill can spawn subagents, write files to the system temporary directory, execute shell commands to open files, and modify local documentation files.
- Sanitization: No sanitization or validation of the ingested codebase content is mentioned. The generated report uses Mermaid with
securityLevel: 'loose', which allows rendering of HTML tags, potentially leading to XSS if the codebase content contains malicious scripts. - [COMMAND_EXECUTION]: The skill instructs the agent to use system commands (
xdg-open,open, orstart) to automatically open the generated HTML report. While this is the primary purpose of the skill, it involves executing shell commands on a file constructed from codebase data. - [EXTERNAL_DOWNLOADS]: The generated architectural report depends on external resources (Tailwind CSS and Mermaid.js) loaded from well-known Content Delivery Networks (CDNs) such as
cdn.tailwindcss.comandjsdelivr.net. These are standard tools for visualization but represent external dependencies in the rendered report.
Audit Metadata