context-fundamentals
Pass
Audited by Gen Agent Trust Hub on Apr 7, 2026
Risk Level: SAFE
Full Analysis
- [DATA_EXFILTRATION]: The scripts/context_manager.py file contains functions to read from the local filesystem to load summaries or detailed documents. This is a documented feature of the context management system and does not target sensitive files or perform unauthorized network operations.
- [PROMPT_INJECTION]: The skill facilitates the ingestion of external data into an agent's context, which is an inherent surface for indirect prompt injection. The documentation provides guidance on using boundary markers and structured delimiters to mitigate this risk. * Ingestion points: ProgressiveDisclosureManager class in scripts/context_manager.py. * Boundary markers: SKILL.md recommends XML tags to delineate context sections. * Capability inventory: Local file read access only; no network access, shell execution, or privilege escalation. * Sanitization: Relies on structural formatting; no explicit code-based sanitization of input strings.
Audit Metadata