The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection (Category 8) because it is designed to ingest and process contents from multiple external files that could be controlled by an attacker or contain unvalidated third-party data.
Ingestion points: The skill reads docs/seo/<slug>/outline.md, docs/seo/<slug>/research-brief.md, docs/seo/brand-voice.md, and docs/seo/topical-clusters.md.
Boundary markers: Absent. The instructions direct the agent to "Internalize All Inputs" and follow the outline "exactly" without using delimiters or system-level instructions to ignore embedded commands within those files.
Capability inventory: The skill has permissions to read various files across the docs/ and references/ directories and write the resulting draft to the local filesystem (docs/seo/<slug>/draft.md).
Sanitization: No sanitization, filtering, or validation is performed on the content read from the input files before it is used to generate the blog post.

seo-write