The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The example configuration in io_config.yaml includes a hardcoded password dev_password for the development profile. While intended for local testing, hardcoding credentials in documentation templates is a suboptimal security practice.
[EXTERNAL_DOWNLOADS]: The skill configuration and deployment examples reference official container images for Mage AI and PostgreSQL.
[COMMAND_EXECUTION]: The documentation provides examples of shell commands using curl to interact with the Mage AI API for management tasks like triggering runs and backfills.
[DATA_EXFILTRATION]: The notify_slack_on_failure.py callback template demonstrates sending pipeline status and error metadata to an external Slack webhook URL.
[PROMPT_INJECTION]: The skill facilitates the creation of pipelines that process data from external sources like Kafka and S3, which introduces a potential surface for indirect prompt injection.
Ingestion points: Data loader blocks fetching records from S3, Kafka, and PostgreSQL (SKILL.md).
Boundary markers: Absent; the provided instructions and code templates do not specify delimiters or warnings to isolate untrusted external content within the agent's context.
Capability inventory: The pipeline blocks possess capabilities for database operations, file system exports, and external network requests via callbacks (SKILL.md).
Sanitization: Absent; the provided code snippets demonstrate data interpolation and processing without explicit validation or escaping of content retrieved from external sources.

mage-ai-pipelines