The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection due to its core function of processing external documents and reviewer comments.
Ingestion points: The skill is designed to read research manuscripts, scientific notes, reviewer comments, and workspace-local instruction files like AGENTS.md (as described in SKILL.md).
Boundary markers: There are no explicit instructions for the agent to use delimiters or sanitization patterns to isolate document content from its internal system instructions.
Capability inventory: The skill instructions permit the agent to edit and create files within the research project's working directory.
Sanitization: No sanitization or validation of the ingested text is performed before it is used to generate or revise publication prose.
[NO_CODE]: The skill consists exclusively of Markdown-based instructions and templates (SKILL.md and referenced .md files). It does not contain executable scripts, package manager configurations, or external code dependencies.

rk-write