The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests data from external websites (Musinsa, Naver, Coupang) which may contain attacker-controlled content. While the skill employs a validation script and a user confirmation gate, it lacks explicit instruction delimiters to prevent the agent from potentially following instructions embedded in scraped product metadata.
Ingestion points: Data fetched via Bright Data scraping tools from third-party e-commerce URLs.
Boundary markers: Absent; the skill does not use specific delimiters or instructions to ignore embedded prompts in the scraped markdown content.
Capability inventory: Writing to local CSV files, executing a local Python validation script, and performing SQL inserts into a Supabase database.
Sanitization: The scripts/validate_csv.py script performs type validation and range checking for numerical fields, which helps ensure data integrity before database insertion.
[COMMAND_EXECUTION]: The skill executes python3 scripts/validate_csv.py as a mandatory step in the data pipeline. This script is bundled with the skill and performs integrity checks on the collected CSV data using standard Python libraries.
[EXTERNAL_DOWNLOADS]: The skill fetches product listing data from external domains including Musinsa, Naver shopping, and Coupang. These operations are core to the skill's purpose of market data collection and use specialized scraping tools.

magma-collect