project-knowledge-base
Pass
Audited by Gen Agent Trust Hub on Feb 20, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- Safe Behavior (SAFE): The provided script
generate_index.pyperforms local indexing of Markdown and JSON files without any network access or high-risk command execution.\n- Indirect Prompt Injection Surface (LOW): The script ingests metadata from user-controlled files and interpolates it into a README.md file, creating a surface where malicious local content could influence an agent.\n - Ingestion points:
generate_index.py(lines 42, 107, 126) reads from local.mdand.jsonfiles.\n - Boundary markers: None; content is directly inserted into the README without delimiters.\n
- Capability inventory: Restricted to local file I/O; no subprocess, eval, or network calls identified.\n
- Sanitization: Metadata strings are not escaped or sanitized before being written to the output file.\n- Missing Files (INFO): Documentation references
init_project_kb.pyandextract_patterns.pywhich are not included in the provided files.
Audit Metadata