The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill relies on reading external local files to establish context and vocabulary, which creates an attack surface for instructions hidden within those files.
Ingestion points: The skill instructs the agent to read domain-modeling/SKILL.md, docs/agents/domain.md, GLOSSARY-MAP.md, GLOSSARY.md, CONTEXT-MAP.md, and CONTEXT.md (as seen in SKILL.md).
Boundary markers: The instructions do not include markers to delimit the external content or explicitly tell the agent to ignore any embedded instructions.
Capability inventory: The skill itself focuses on stylistic adjustment (Simplified Technical English) and does not call external tools or execute commands.
Sanitization: There is no evidence of content sanitization or validation for the files being read into context.

wait-what