The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted input from source files and user notes to synthesize wiki articles, which could theoretically contain instructions to manipulate the agent's curation logic.
Ingestion points: Content is ingested from the inbox/ and raw/ directories, as well as loose user notes (SKILL.md).
Boundary markers: The instructions do not specify the use of clear delimiters or guardrail prompts when interpolating untrusted data into the gardening process.
Capability inventory: The skill's capabilities are limited to reading and writing files within the 02-DOCS/ wiki structure; it lacks network access, shell execution, or sensitive system file access.
Sanitization: The skill enforces a high bar for capture (defaulting to discard) and uses specific markdown templates, which limits the influence of external data.
[SAFE]: The skill is well-scoped to documentation management and incorporates several best practices for data integrity, such as preservation before overwrite and append-only/prepend-only audit logs.

knowledge-ops