The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from local files (CSV, XLSX, TXT) to perform enrichment lookups.
Ingestion points: The load_rows function in scripts/enrich.py reads data from user-supplied input files.
Boundary markers: Data is parsed into structured dictionaries based on column headers before being passed to the API request builders.
Capability inventory: The script performs network requests to the PDL API and writes structured results to .xlsx files using openpyxl.
Sanitization: Input values used in SQL-like query strings are sanitized via the _esc function, which lowercases, trims, and escapes single quotes to prevent query injection against the PDL API.
[EXTERNAL_DOWNLOADS]: The skill communicates with the People Data Labs API to fetch enrichment data.
Source: Requests are sent to api.peopledatalabs.com.
Purpose: This is the primary intended function of the skill (enrichment and search). Requests are made using the standard urllib library with a fixed HTTPS base URL.
[CREDENTIALS_UNSAFE]: The skill provides clear instructions and code for handling API keys securely.
Mechanism: The resolve_api_key function in scripts/enrich.py retrieves the PDL_API_KEY from environment variables or a local .env file.
Security Practice: The documentation explicitly warns against hardcoding or committing the API key to version control, which aligns with industry best practices for secret management.
[DATA_EXFILTRATION]: While the skill transmits PII (names, companies, emails) to a third party, this is the explicit and documented purpose of the enrichment service.
Context: The skill does not access sensitive system files (e.g., SSH keys, AWS credentials). The transmitted data is limited to the inputs provided by the user for enrichment purposes.

people-enrichment