The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill facilitates the ingestion and processing of external data files (CSV, JSON, and Parquet) through SQL queries and file listings. This creates a potential surface where malicious instructions embedded in stored files could influence agent logic.
Ingestion points: Data retrieved from Celigo Storage via file listings or SQL queries (SKILL.md, references/storage-query.md).
Boundary markers: The skill does not define specific delimiters for isolating external data within the agent's prompt context.
Capability inventory: The skill possesses extensive file management capabilities (upsert, move, copy, merge, delete), SFTP account administration, and SQL execution.
Sanitization: The Storage Query service enforces a strict allowlist of read-only operations (SELECT, DESCRIBE, and SUMMARIZE) and applies an account-wide rate budget.
[CREDENTIALS_UNSAFE]: The documentation addresses the handling of sensitive credentials, specifically presigned S3 URLs and SFTP partner passwords. It includes mandatory security instructions to treat these as temporary, avoid logging or storing them, and use secure CLI input methods (stdin) to prevent password leakage in process logs or chat history.
[COMMAND_EXECUTION]: The skill provides numerous CLI commands for the celigo tool to manage platform resources. These commands are legitimate administrative functions used for the primary purpose of storage and file server management.

managing-file-storage