last30days
Pass
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from Reddit, X, and web sources (documented in
scripts/lib/openai_reddit.py,scripts/lib/xai_x.py, andscripts/lib/websearch.py) to synthesize research and generate prompts for other AI tools. Boundary markers are absent; while the skill uses Markdown for structure, it lacks explicit instructions to the agent to disregard potentially malicious directives embedded within the fetched content. The skill's capability inventory includesexec(running the research script),file_write(to the skill's local configuration), andnetwork(accessing social media APIs and general web content). Sanitization is implemented via text truncation in the library modules. - [COMMAND_EXECUTION]: The skill utilizes the
execpermission to run a local Python script (scripts/last30days.py) that handles API communication, data processing, and report generation.
Audit Metadata