The Agent Skills Directory

[SAFE]: The skill performs transparent file system operations to create a directory structure and markdown files. No malicious code, obfuscation, or unauthorized access patterns were identified.
[PROMPT_INJECTION]: The skill establishes an indirect prompt injection surface by defining how the agent should ingest and synthesize information from external files (e.g., in the raw/ directory or project README files). Content in these sources could contain instructions designed to manipulate the agent's output or suggestions.
Ingestion points: The raw/ directory for source documents and codebase files such as README.md and dependency manifests.
Boundary markers: The generated SCHEMA.md and codebase.md provide structure, frontmatter requirements, and specific rules for handling README content to separate design intent from operational details.
Capability inventory: The agent is instructed to write wiki pages and is empowered to evaluate and suggest edits to project READMEs based on the ingested content.
Sanitization: There are no explicit instructions to sanitize source content or to ignore potential instructions embedded within the processed text.

wiki-init