The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses bash and python3 to execute a search script (scripts/search_google_events.sh) which performs network requests and handles JSON processing. User-provided arguments for the search query and filters are passed to this script.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from the SerpAPI Google Events response through Frevana's endpoint.
Ingestion points: External data enters the agent context via the stdout of the search_google_events.sh script, which prints the raw JSON response from https://ai-factory.frevana.com/service/google-events.
Boundary markers: The skill body advises the agent to "not rewrite or reshape returned data unless the user explicitly asks," but no explicit instruction-level delimiters are used when the agent reads the search results.
Capability inventory: The skill has the capability to write the results to a local file using the --output flag in scripts/search_google_events.sh and perform POST network requests via curl to the Frevana API.
Sanitization: The Python helper scripts validate that the input arguments match expected types (int, bool, string) and ensure the final API response is valid JSON before outputting, which mitigates simple schema confusion but does not filter the natural language content of the event results.
[DATA_EXFILTRATION]: The skill uses an environment variable FREVANA_TOKEN or a command-line argument --token to handle Bearer authentication for the API request. While it correctly suggests storing secrets in the environment, the command execution pattern reveals these secrets to the local process list if passed via --token.

google-events-search