last30days
Pass
Audited by Gen Agent Trust Hub on Mar 22, 2026
Risk Level: SAFE
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from Reddit, X, and WebSearch. While this introduces an inherent surface for indirect prompt injection (where malicious instructions could be embedded in search results), the skill mitigates this by using structured markdown headers, distinct output sections, and explicit instructions for the 'Judge Agent' to identify patterns and ground synthesis in actual content.
- Ingestion points: Fetches content from Reddit threads, X posts, and general web pages.
- Boundary markers: Uses clear markdown headers (e.g., '### Reddit Threads') and separators ('---') in the generated reports.
- Capability inventory: Limited to internal data synthesis and generating prompt text; no arbitrary command execution or unrestricted file access on untrusted input.
- Sanitization: Implements text normalization, deduplication, and date-based filtering to ensure relevance and structure.
- [CREDENTIALS_SAFE]: The skill implements standard secret management practices for CLI tools. It instructs users to store API keys in a dedicated configuration file (~/.config/last30days/.env) with restricted permissions (chmod 600) and provides a setup check to verify key availability without hardcoding secrets.
- [COMMAND_EXECUTION]: The skill uses local Python scripts for orchestration and data processing. These scripts are invoked via the shell but operate strictly within the skill's own directory structure and standard system paths (~/.local/share, ~/.cache, ~/.config) for state management.
Audit Metadata