version: "1.0.2"

Skill: Python RAG Audit

Mission

Audit Python RAG (Retrieval-Augmented Generation) pipelines for security, data leakage, prompt injection risks, and production readiness.

Allowed Actions

• Read Python source files.
• Read requirements and dependency files.
• Analyze RAG pipeline code.
• Detect prompt injection vulnerabilities.
• Detect PII/PHI exposure in data processing.
• Detect hardcoded API keys and secrets.
• Analyze embedding and vector store configurations.
• Generate audit reports.

Forbidden Actions

• Modify production data.
• Access external APIs without approval.
• Read secret values.
• Execute model inference.
• Disable safety guardrails.

Inputs

• workspace path
• pipeline entry points
• data source paths

Outputs

• rag-security-report.md
• dependency-vulnerability-report.md
• data-exposure-report.md
• hardening-recommendations.md

Evidence Required

• pipeline code with line references
• dependency list with versions
• data processing flow
• vector store configuration
• prompt template citations

Prompt Contract

• State the objective, target scope, assumptions and constraints before execution.
• Use only evidence-backed facts; mark uncertainty explicitly.
• Route tool access through approved command, MCP or Tool Router paths.
• Redact secrets, credentials, tokens and sensitive values.
• Return facts, assumptions, risks, recommendations, evidence_refs and blocking_conditions when applicable.
• Stop when required evidence, permissions, policy approval or input context is missing.

Quality Gates

• Must identify each vulnerability with code location.
• Must classify severity with rationale.
• Must not expose actual data samples in report.
• Must provide actionable fix recommendations.