The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted technical prose and documentation provided by the user for translation, editing, and quality assurance.
Ingestion points: Technical articles, prose drafts, and documentation sections are ingested as input for translation and review tasks (SKILL.md, references/english-to-japanese-workflow.md).
Boundary markers: The prompt templates provided in the reference documentation use open placeholders such as [Paste section.] and [Paste source.] without employing strict delimiters or providing explicit instructions to the agent to disregard instructions potentially embedded within the user-supplied text (references/english-to-japanese-technical-translation.md).
Capability inventory: The documentation recommends the use of shell-based linting tools such as vale and harper-cli on local documentation files (references/vale-and-harper.md).
Sanitization: No evidence of sanitization or validation logic is present to inspect user-provided content for malicious instructions before interpolation into model prompts.

technical-writing