The Agent Skills Directory

[PROMPT_INJECTION]: The skill is designed to ingest and process natural language prose from the external mattpocock/skills repository, creating a surface for indirect prompt injection.
Ingestion points: Processes upstream content including READMEs, skill instructions, docs, and frontmatter prompts from SKILL.md files.
Boundary markers: The skill contains explicit rules to protect behavior-critical spans, preserve Markdown structure, and uses a 'Fail-closed' rule to ensure ambiguous text is preserved rather than rewritten.
Capability inventory: The agent is instructed to read/write files in the local repository and execute node-based scripts.
Sanitization: The instructions strictly forbid translating code blocks, CLI commands, file paths, and identifiers, which mitigates the risk of accidental execution of injected commands within translated output.
[COMMAND_EXECUTION]: The skill instructs the agent to execute specific local maintenance and validation scripts.
Evidence: Instructions specify running node scripts/check-translation.mjs and node scripts/audit-english.mjs to verify structural integrity and identify English terms for review.

translate-skill