The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests and processes raw material from user-specified file paths to create project briefs. This data ingestion process provides an attack surface for indirect prompt injection if those files contain malicious instructions meant to influence the agent's summaries or decisions.
Ingestion points: External data sources and local project files (e.g., context/foundation/, DESIGN.md) identified in the input arguments.
Boundary markers: Instructions within the skill explicitly differentiate between raw data and executable instructions, specifically using the untrusted-content parameter to limit the influence of fetched material.
Capability inventory: The skill is capable of reading/writing files in specific project subdirectories, calling other discovery tools (dx-references, dx-domain), and spawning subagents for review.
Sanitization: Includes a mandatory 'skeptic pass' to validate claims against source material and utilizes specialized tool modes designed to neutralize instructions found in untrusted data.

dx-distill