The Agent Skills Directory

[DYNAMIC_EXECUTION]: The script scripts/search_google_trends.sh employs a heredoc pattern (`python3
<<'PY'`) to execute embedded Python code at runtime. This method is used twice: once to construct the JSON request payload from environment variables and once to validate that the API response is valid JSON before saving it to disk.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process data from an external API (https://ai-factory.frevana.com). Because the agent is instructed to summarize this data, it is vulnerable to instructions or malicious patterns embedded within the search trend results that could attempt to influence the agent's behavior.
Ingestion points: scripts/search_google_trends.sh via curl request to the Frevana API.
Boundary markers: None present in the current implementation to separate data from instructions.
Capability inventory: The skill uses curl for network requests and python3 for file and JSON processing.
Sanitization: The skill validates that the output is JSON but does not sanitize the content of the data for potential prompt injection before it is provided to the agent.
[DATA_EXPOSURE_AND_EXFILTRATION]: The skill requires a FREVANA_TOKEN (Bearer token) for authentication. While the script includes a note not to echo the token, it is passed via command-line arguments and environment variables, which may be visible in process listings or logs. Additionally, search queries are sent to the external domain ai-factory.frevana.com.

google-trends