The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from external AI model responses and web search results, creating a potential surface for indirect prompt injection.
Ingestion points: The skill reads model outputs and cited page content via API calls and search tools, storing them in runs.jsonl in the local filesystem.
Boundary markers: Includes an explicit instruction to the agent to treat answer text and cited pages strictly as data and to ignore any instructions embedded within them.
Capability inventory: The skill has the ability to perform network operations, write to the filesystem, and execute a local Python script.
Sanitization: Data is parsed as JSON and matched using regular expressions in a bundled script, but there is no deep structural sanitization of the natural language content itself.
[COMMAND_EXECUTION]: The skill executes a bundled Python script (scripts/tally.py) to process logged data and calculate visibility metrics. This script is distributed with the skill and uses only the Python standard library.
[EXTERNAL_DOWNLOADS]: The skill interacts with external AI provider APIs (OpenAI, Perplexity, and Gemini) using curl to fetch response data. It also makes a request to the vendor's own domain (revnu.com) to provide preview functionality for users who have not yet connected the full service.

revnu-geo-check