The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external patent data (titles, abstracts, and technical summaries) from files such as tagged_pool.csv.
Ingestion points: Data enters the context via tagged_pool.csv, panorama_stats.json, and value_signals.json as defined in the input requirements of SKILL.md.
Boundary markers: The skill includes specific instructions to perform 'pure file operations' using scripts for bulk exports to avoid reading large numbers of data rows directly into the conversational context.
Capability inventory: The skill generates HTML reports and CSV/JSON files in the @session/pps-output/ directory and invokes Patsnap MCP tools for semantic text analysis.
Sanitization: The skill relies on the structural integrity of the input data but does not explicitly mention sanitization or filtering of technical text fields.

pps-tag