The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to read and process potentially untrusted external data sources, such as project instructions and brand documents, to guide the AI agent's implementation and testing tasks. This creates a surface for indirect prompt injection attacks.
Ingestion points: In SKILL.md (Step 1), the agent is instructed to "Read project instructions, product and brand documents, the relevant route and components" to identify the user's job and constraints.
Boundary markers: The skill does not define specific delimiters or instructions to prevent the agent from obeying malicious commands that might be embedded within these external documents.
Capability inventory: The agent has the capability to generate and implement web code (Tailwind v4), execute the application locally, and perform interactive browser-based QA to exercise user paths and verify persistence (SKILL.md Step 5).
Sanitization: The skill contains no instructions for sanitizing or filtering the content of the external documentation before it is integrated into the agent's reasoning process.

chiaroscuro