The Agent Skills Directory

[COMMAND_EXECUTION]: The skill is designed around the italianparliament CLI and encourages the agent to use shell pipelines involving jq, grep, wc, and duckdb. It also instructions the agent on how to redirect output to files (e.g., > bills-19.csv).
[EXTERNAL_DOWNLOADS]: The bill-text fetch command downloads external documents (PDFs/HTML) from official parliament websites (www.senato.it) and requires external tools like agent-browser and lit to process them. It also interacts with external SPARQL endpoints at camera.it and senato.it.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection and command injection. It instructs the agent to interpolate user-provided strings (like names, keywords, and search terms) directly into shell commands (e.g., italianparliament search find --name <user_input>).
Ingestion points: User-supplied names, keywords, URIs, and search terms in SKILL.md.
Boundary markers: None present; parameters are passed directly to CLI flags.
Capability inventory: Execution of shell commands via italianparliament, shell pipes, file system writes, and browser automation via agent-browser (noted in references/commands.md).
Sanitization: The instructions do not include guidelines for escaping or validating user input before command construction.

italian-parliament-cli