The Agent Skills Directory

[SAFE]: The skill provides automated utilities for engineering tasks without requesting dangerous permissions or performing suspicious operations. The Python scripts are written using only standard library modules and perform safe JSON and text processing.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted JSON data provided by users (such as database schemas and migration specifications) to generate reports and plans. This is a common architectural surface where malicious input data could attempt to influence the agent's behavior.
Ingestion points: The scripts compatibility_checker.py, migration_planner.py, and rollback_generator.py ingest user-supplied JSON files via command-line arguments.
Boundary markers: The generated human-readable reports and plans do not use specific delimiters or instructions to the LLM to ignore embedded content from the input data.
Capability inventory: The Python scripts are limited to file system access and text generation. They do not perform network requests or execute shell commands.
Sanitization: Standard JSON parsing is used, but the resulting report strings are not sanitized for potential prompt injection markers before being presented to the agent.

migration-architect