The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by treating prose rules inside a configuration file as executable logic for the agent.
Ingestion points: The skill reads instructions directly from ~/Obsidian/CONVENTIONS.md and ~/Obsidian/CONTEXT.md at runtime.
Boundary markers: None identified; the skill instructions describe the prose rules as "directly executable" for the agent.
Capability inventory: The skill has the ability to read and write files within the vault, including performing batch renames and updating backlink content.
Sanitization: There is no evidence of filtering or validation of the content within the rule files before the agent interprets them.
[COMMAND_EXECUTION]: The skill performs automated file system operations, including file creation, modification of frontmatter, and renaming files with atomic link updates, based on its interpretation of the rule set.

lint