rcca-master
Pass
Audited by Gen Agent Trust Hub on Oct 30, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user input from problem descriptions and customer complaints, which serves as a potential surface for indirect prompt injection. Ingestion points: User-provided text enters the agent's context during the 8D workflow phases and is passed to the report generation scripts. Boundary markers: The SKILL.md file includes explicit behavioral guardrails, directing the model to treat all user text as data and to ignore instruction-like content within inputs. Capability inventory: The scripts scripts/initialize_8d.py and scripts/generate_8d_report.py perform local file-write operations; no network access, subprocess execution, or dynamic code evaluation capabilities are present. Sanitization: The scripts/generate_8d_report.py script utilizes the html.escape() function to sanitize user data before it is embedded in HTML reports, mitigating Cross-Site Scripting (XSS) risks.
- [SAFE]: The Python scripts implement secure path handling by resolving absolute paths, checking for traversal patterns, and enforcing strict file extension allowlists. No evidence of obfuscation, exfiltration, persistence, or malicious command execution was found.
Audit Metadata