The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: Fetches parliamentary session transcripts and official journal data from the author's API at parlamento.ai.
[COMMAND_EXECUTION]: Uses curl to interact with data endpoints and to send generated HTML content to a PDF rendering service hosted on Google Cloud Run (source-worker-876875904047.us-central1.run.app).
[DATA_EXPOSURE]: Uses an environment variable $PARLAMENTO_API_KEY for authorization, which is the standard secure method for handling credentials in this environment.
[PROMPT_INJECTION]: The skill processes external parliamentary data, which presents an attack surface for indirect prompt injection. However, the skill implements a strict multi-phase validation process (Source Inventory and Metrics Validation) to ensure the agent's analysis remains grounded in the provided data.
Ingestion points: Parliamentary transcripts and Official Journal data fetched from https://parlamento.ai/api/external/research/* (Phase 3).
Boundary markers: Absent in the prompts.
Capability inventory: Local file writing (Write tool to /tmp/report.html) and network requests via curl to transmit data to the PDF generator.
Sanitization: Absent; content is analyzed by the LLM and formatted into a template before PDF generation.

research