project-onboarding
Pass
Audited by Gen Agent Trust Hub on Sep 20, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill implements a workflow for ingesting potentially untrusted external data into a RAG (Retrieval-Augmented Generation) system, creating an attack surface for indirect prompt injection.
- Ingestion points: Documentation files including README.md, ARCHITECTURE.md, DESIGN.md, CONTRIBUTING.md, and all Markdown files within the docs/ directory are read and ingested into RAG collections (
scripts/onboard-docs.sh). - Boundary markers: No boundary markers, delimiters, or system instructions are used to separate the ingested content from the agent's primary instructions or to warn the model about untrusted content.
- Capability inventory: The skill possesses capabilities for writing configuration files (
config/projects/), creating filesystem symlinks (ln -sf), and executing shell and Python scripts (scripts/onboard-project.sh). - Sanitization: There is no evidence of validation, filtering, or sanitization of the file contents before they are stored in the RAG collections.
Audit Metadata