The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted documents from the raw/ directory to build the wiki, which presents an attack surface for indirect prompt injection. Maliciously crafted source documents could influence the agent's behavior during ingestion or querying.
Ingestion points: Files in the raw/ directory read during the Ingest and Query operations defined in SKILL.md.
Boundary markers: The skill uses YAML frontmatter and structured Markdown but lacks explicit instructions to ignore or delimit embedded commands within source documents.
Capability inventory: File system read/write operations (Markdown and logs) in SKILL.md and local shell execution in scripts/init-wiki.sh.
Sanitization: No specific sanitization, filtering, or validation of the content within ingested source documents is mentioned.
[COMMAND_EXECUTION]: The skill includes a bash script scripts/init-wiki.sh used to initialize the wiki's directory structure and initial files.
The script uses standard shell utilities such as mkdir, cat, and sed to operate on the local file system.
While standard for setup, the script accepts user-provided arguments for the wiki name and base directory.

llm-wiki