The Agent Skills Directory

[SAFE]: The skill is a workflow orchestrator that manages scientific writing tasks. It does not contain executable code, shell commands, or network operations.
[PROMPT_INJECTION]: The skill ingests data from a repository to generate text artifacts. 1. Ingestion points: repository files (code and results), story.md, and outline.md. 2. Boundary markers: No explicit delimiters or instructions to ignore embedded commands are present. 3. Capability inventory: The skill routes to sub-skills for writing tasks but does not directly execute shell commands or perform network operations. 4. Sanitization: The workflow is gated by mandatory manual user approval in the conversation before advancing to the next stage, which provides a human-in-the-loop security boundary against indirect prompt injection.

scientific-writing