The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill defines a surface for indirect prompt injection by instructing the agent to interpolate content from a local file.
Ingestion points: The agent is instructed in SKILL.md to use definitions and language from CONTEXT.md.
Boundary markers: No explicit delimiters or instructions to ignore embedded commands are provided.
Capability inventory: The skill restricts the agent's capabilities by setting disable-model-invocation: true in the SKILL.md frontmatter, preventing tool execution or network access.
Sanitization: No sanitization or filtering of the external file's content is specified.

wait-what