The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill contains multiple 'auditor' modules (e.g., codable-auditor.md, swiftdata-auditor.md, storage-auditor.md) designed to instruct an AI agent to perform security and best-practice scans on a project's codebase.
Ingestion points: The auditor instructions direct the agent to use Glob, Grep, and Read tools to ingest arbitrary Swift source files (**/*.swift) and data model files (**/*.xcdatamodeld) from the user's environment into the agent's context.
Boundary markers: The audit routines lack explicit boundary markers or instructions to disregard potential prompt injections embedded within the user's source code being analyzed.
Capability inventory: The skill leverages powerful analysis capabilities, including Grep and Read tools, to process and interpret the ingested code.
Sanitization: No explicit sanitization or filtering is performed on the content read from the user's project before it is processed by the agent.

axiom-data