The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data by reading markdown frontmatter tags from vault files to generate a configuration array.
Ingestion points: Markdown frontmatter tags field parsed from all .md files in the vault (SKILL.md, Step 2).
Boundary markers: The instructions do not specify any delimiters or warnings to ignore embedded instructions within the parsed tags.
Capability inventory: The skill performs file read operations across the vault and file write operations to .obsidian/graph.json, backup files, and log.md.
Sanitization: There is no explicit requirement for the agent to sanitize or escape tag strings before interpolating them into the colorGroups JSON query field.

graph-colorize