The Agent Skills Directory

[SAFE]: The skill is composed exclusively of markdown instructional files and a standard MIT license. It does not contain executable scripts, configuration files that modify system state, or commands that perform network operations.
[INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection as it is designed to process and edit untrusted, user-supplied text. However, the instructions include specific mitigations and safety protocols:
Ingestion points: Source text and writing samples provided by the user (noted in SKILL.md and references/patterns.md).
Boundary markers: The skill explicitly instructs the agent to "Preserve what the evidence says and the writer actually means" and "Repair unsupported additions," creating a conceptual boundary between user data and editorial instructions.
Capability inventory: The skill uses the internal skill://coder tool for variant generation but restricts its use to "exact authorized provider and model" with "independent input packets."
Sanitization: The manual review process (Section 4 of SKILL.md) and the "Review Contract" (references/review-contract.md) provide a verification layer to ensure that the output has not been corrupted by input instructions.

writing