The Agent Skills Directory

[SAFE]: The skill's instructions and example workflows are consistent with its stated purpose of data exploration. It relies on a set of purpose-built tools for interacting with the Honeydew semantic layer.
[PROMPT_INJECTION]: No attempts to override agent behavior, bypass safety guardrails, or extract system prompts were found. The instructional content is professional and focused on the intended data analysis tasks.
[DATA_EXFILTRATION]: There is no evidence of sensitive data harvesting or unauthorized network transmission. The skill uses a session-bound workspace and branch context, and all data retrieval occurs through authenticated semantic layer tools.
[INDIRECT_PROMPT_INJECTION]: The skill possesses an inherent attack surface for indirect prompt injection as it ingests and processes data from external sources (the semantic layer).
Ingestion points: Data rows returned by get_data_from_fields and analysis reports generated by ask_deep_analysis_question.
Boundary markers: The skill does not define specific delimiters (e.g., XML tags or backticks) to wrap retrieved data or instruct the agent to ignore instructions embedded within the data.
Capability inventory: The skill is restricted to model exploration and querying; it lacks dangerous capabilities such as file system writes, shell execution, or unrestricted network operations.
Sanitization: No sanitization or validation logic for the retrieved data is mentioned in the instructions.

query