About this skill

What problem it solves

Data analysts often juggle local scripts, database credentials, API tokens, and business rules. Handing scripts or secrets directly to an AI model can leak sensitive content into the prompt context, while storing scripts in plaintext makes collaboration and permission boundaries harder to manage. This suite is not designed to let an AI automatically read data or perform analysis. Instead, it keeps sensitive assets inside a local security boundary and lets an Agent use credentials, apply scripts, or read knowledge only after explicit user authorization, reducing the chance that script or credential content reaches the model.

How it works

The skill is primarily built from prompt documents in prompts/, and it can also use agent_system_prompt.md to configure a data analyst Agent. It organizes work around three asset types:
- Credential management: Encrypt and store sensitive credentials such as database keys and API tokens; scripts can call them with zero exposure, while the Agent should not read plaintext values.
- Script management: Encrypt user-owned query, cleaning, and analysis scripts; apply them through mgc_run or mgc_get after authorization; use mgc_seal when scripts need collaborative sharing while preserving authorization boundaries.
- Knowledge management: Preserve frameworks, prompt templates, methodologies, and business rules so the Agent can read them under authorization and shape analysis structure without auto-editing scripts or auto-fetching data.

A typical flow is: the user supplies scripts and credentials, the Agent requests authorization for each sensitive action, then calls MGC MCP tools to save, search, apply, or seal entries. mgc_find can fuzzy-locate entries and mgc_list can list them exactly, while sensitive operations still require explicit consent.

Boundaries and notes

The suite does not provide automatic data access, automatic cleaning, automatic transfer, or script generation and modification. All scripts must be user-provided and compliant with organizational policy; the AI should only receive execution results, not script or credential content. It fits internal data analysis, secure use of vendor scripts, team collaboration, and knowledge capture, but it is not for Agents that autonomously choose data paths or rewrite analysis logic.

Screenshots

Use Cases

Data analysts need to store database secrets locally and let scripts call them without exposing credentials.
Engineers need to share proprietary cleaning scripts with a collaboration node while keeping them encrypted and sealed.
Teams need to preserve analysis frameworks, prompt templates, and business rules, then let an Agent read them under authorization.
Data teams receiving vendor query scripts need to ensure the AI only receives execution results, not script or credential plaintext.

Best For

Data analysts managing database credentials and API tokens who need zero-exposure script access
Engineers who need to share sealed query scripts without exposing plaintext code
Data teams that want to reuse analysis frameworks and prompt templates
Platform engineers building data analyst Agents with strict authorization boundaries