The Agent Skills Directory

Data Exposure & Exfiltration (MEDIUM): The skill is specifically designed to aggregate application logs (customerio-debug.log) and API responses into a compressed bundle.
Evidence: The scripts collect userId and event data (attributes) which typically contain PII like email addresses, names, and phone numbers.
Mitigation: Although the skill uses sed to redact api_key patterns, it does not sanitize PII or other sensitive customer data before bundling.
Indirect Prompt Injection (LOW): The skill has an ingestion surface for untrusted data by reading application logs and API outputs.
Ingestion points: customerio-debug.log, api-test.json (File system reads).
Boundary markers: Absent (no explicit delimiters or warnings for downstream agents).
Capability inventory: Bash(curl:*), Read, Grep across multiple scripts.
Sanitization: Redaction is limited to API keys; payload content remains raw.
Command Execution (SAFE): The skill uses curl, npm, and pip for connectivity tests and environment discovery, which aligns with its primary diagnostic purpose.

customerio-debug-bundle