The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from the Obsidian vault (markdown files) and interpolates it into an interactive HTML visualization (graph.html).
Ingestion points: Reads frontmatter (id, label, category, tags, summary) and body content (wikilinks) from all .md files in the vault.
Boundary markers: Absent. The instructions do not define delimiters or warnings to prevent the agent from being influenced by malicious content embedded in the notes.
Capability inventory: The skill writes multiple files to the local file system (graph.json, graph.graphml, cypher.txt, postgres.sql, graph.html, and an OKF bundle).
Sanitization: Although the SQL generator mentions doubling single quotes for escaping, the interactive HTML visualization script directly assigns node metadata to innerHTML without escaping, which could execute malicious scripts if the vault content is compromised.
[DYNAMIC_EXECUTION]: The skill generates multiple files intended for execution or interpretation in other environments.
It creates postgres.sql and cypher.txt containing database commands for upserting the graph data.
It creates graph.html containing JavaScript and a reference to the vis-network library (via the unpkg.com CDN) to render the graph at runtime in a browser.
[EXTERNAL_DOWNLOADS]: The generated graph.html file includes a script tag fetching the vis-network library from unpkg.com. This is a well-known service used for legitimate interactive visualization in the skill's output.

wiki-export