The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to process and analyze user-supplied code, modules, and Pull Requests, which serves as an ingestion point for untrusted data. This creates a vulnerability where malicious instructions embedded in the target code could influence the agent's diagnostic findings or behavior.
Ingestion points: SKILL.md identifies the "file, module, or PR to review" as the input target for the analysis funnel.
Boundary markers: No explicit delimiters or "ignore embedded instructions" directives are specified for the content read from the target files.
Capability inventory: The skill uses Read and Grep tools to ingest data and can spawn multiple sub-agents to perform structural and interface reviews.
Sanitization: The instructions do not contain logic for sanitizing or validating the content of the files before they are processed by the agent or sub-agents.
[EXTERNAL_DOWNLOADS]: The skill describes itself as an orchestrator for the "Clairvoyance" collection of skills (referenced at https://clairvoyance.fyi). It relies on these external, non-standard components being present and installed in the agent's environment to execute its full diagnostic funnel.
[DYNAMIC_EXECUTION]: The references/workflow-builder.md file provides a framework for generating "Dynamic Workflow" scripts at runtime. These scripts use pipeline and parallel constructs to dynamically coordinate multiple agents and handle results based on the analysis targets, representing a complex orchestration layer that evolves during the session.

design-review