The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted external data, including existing drafts, rubric files, and source evidence. This creates a surface for indirect prompt injection attacks where malicious instructions could be embedded within the academic text processed by the agent.
Ingestion points: The skill reads existing_draft, rubric_or_constraint, and source_data as specified in the Expected inputs and Logical Operations sections of SKILL.md.
Boundary markers: The skill contains a protocol to quote literal text from supplied rubrics before applying them, which provides some context separation but does not prevent adversarial instruction following.
Capability inventory: The skill possesses Read, Edit, and Write capabilities on the LOCAL_FS to modify files and generate prose artifacts.
Sanitization: There are no documented procedures for sanitizing or escaping the user-provided text before it is incorporated into the agent's context.

oma-academic-writing