stock-copilot-pro
Pass
Audited by Gen Agent Trust Hub on Oct 2, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes news headlines and social media posts from external market data providers. While this presents a theoretical surface for indirect prompt injection, the risk is low due to the structured nature of the analysis and the use of protective measures like fenced code blocks for raw data and evidence sections. * Ingestion points: News and social media content are fetched via qveris.ai APIs and parsed in scripts/lib/data/parser.mjs. * Boundary markers: Analysis results are presented to the agent using clear Markdown headers and structured tables. * Capability inventory: The skill is restricted to outbound HTTPS calls to qveris.ai and local file writes to its own config and .evolution directories. * Sanitization: A comprehensive redactSecrets function masks API keys and tokens in outputs, and toFencedCodeBlock ensures raw content is treated as text.
- [DATA_EXPOSURE]: The skill manages a local watchlist and runtime evolution state. It proactively redacts sensitive environment variables (QVERIS_API_KEY) and authorization headers from all logs, reports, and persisted state files to prevent credential exposure.
- [EXTERNAL_DOWNLOADS]: The skill fetches data and configuration from qveris.ai. This is the official domain for the skill's data services and is explicitly allowed in the skill's network configuration. The script enforces host-validation to prevent requests to unauthorized domains.
Audit Metadata