Skills
Navigate
skills/svenja-dev/claude-code-skills/project-development/Gen Agent Trust Hub

project-development

Pass

Audited by Gen Agent Trust Hub on Sep 27, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill documents the creation of LLM pipelines that ingest untrusted external data (e.g., from APIs, web scraping, or Hacker News comments). \n
  • Ingestion points: Data is acquired in stage_acquire (scripts/pipeline_template.py) and processed in the fetch stages described in the case studies (references/case-studies.md). \n
  • Boundary markers: Prompt templates in SKILL.md and scripts/pipeline_template.py use structural delimiters (e.g., ---, ## Section) and explicit instructions for the model to follow a specific format. \n
  • Capability inventory: The provided template performs file system writes to local directories to store intermediate and final results. The case studies mention broader capabilities in production environments like ExecuteCommand (bash). \n
  • Sanitization: The render_html function in scripts/pipeline_template.py implements html.escape for final HTML rendering to mitigate potential XSS vulnerabilities from processed data.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 27, 2026, 09:07 PM