The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the Snowflake CLI (snow sql) to collect pipeline metadata. This is implemented securely in collect_snowflake_evidence.py by using parameterized command lists and strictly validating CLI profile names against a safe regular expression.- [INDIRECT_PROMPT_INJECTION]: The skill processes data returned from Snowflake queries. It mitigates injection risks by enforcing a strict schema contract, redacting potential credentials via regular expressions, hashing account-scoped identifiers (pseudonymization), and validating data values against fixed enums before analysis.- [SAFE]: Static analysis identified a potential prompt injection in eval-spec.yaml. This string is a legitimate test case within an evaluation harness designed to verify the agent's resistance to such attacks, and does not represent a malicious attempt by the skill itself.- [SAFE]: The SQL templates used for data collection are verified to be read-only. The collection script explicitly checks for and rejects SQL containing mutation keywords (e.g., ALTER, CREATE, DROP, INSERT) and unapproved system functions.

snowflake-pipeline-guardian