The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from user-controlled markdown files within the Obsidian vault to generate new synthesis pages. This creates a surface for indirect prompt injection where malicious instructions hidden in the wiki could influence the agent's behavior during the synthesis process.
Ingestion points: The skill scans all non-special .md files in the vault, specifically extracting [[wikilinks]], tags, and category frontmatter.
Boundary markers: No explicit delimiters or instructions to ignore embedded commands are defined for the data extracted from source pages.
Capability inventory: The skill has the ability to execute shell commands (rg, obsidian-wiki, qmd) and write new markdown files to the synthesis/ directory.
Sanitization: There is no evidence of sanitization or escaping of the content extracted from the vault before it is used to generate new titles, summaries, or insights.
[COMMAND_EXECUTION]: The skill uses shell commands to interact with the system and vault.
Evidence: Usage of rg (ripgrep) for searching, obsidian-wiki for state synchronization and task management, and qmd for vector index updates. These tools are used within their intended functional scope for vault management.

wiki-synthesize