The Agent Skills Directory

[PROMPT_INJECTION]: The skill uses the $ARGUMENTS variable directly within its instructions. This creates a surface for direct prompt injection where a user could provide input designed to override the agent's programmed behavior. Furthermore, the skill processes local codebase files which could contain adversarial content designed to influence the agent's output.
Ingestion points: The skill retrieves content from the codebase using Read, Grep, and Glob tools based on the user's input.
Boundary markers: Absent. There are no instructions or delimiters defined to separate the skill's instructions from the content of the files it analyzes.
Capability inventory: The skill uses Read, Grep, and Glob tools as defined in the frontmatter.
Sanitization: Absent. Data read from the filesystem is not sanitized before being processed by the agent.
[SAFE]: The skill explicitly enforces safety protocols, specifically stating that it must not execute SQL commands, connect to production databases, or design schemas from scratch. It prioritizes risk analysis and documentation over direct action.

db-migration-draft