The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user-controlled translation files (.md, .tex, .txt) which could contain malicious instructions designed to influence the agent's parity report or summary behavior.
Ingestion points: The skill reads chapter files and sibling translations across a user-defined directory layout, as well as project-specific story bibles (SKILL.md, references/checklist.md).
Boundary markers: There are no explicit delimiters or instructions defined to treat audited prose as data only, increasing the risk that the agent may follow instructions embedded within the book content.
Capability inventory: The skill utilizes Bash, Read, Grep, and Glob tools to perform its auditing functions and execute Git commands (SKILL.md).
Sanitization: The instructions do not specify any validation or sanitization steps for the file content before it is analyzed and processed by the agent.

translation-sync