The Agent Skills Directory

[SAFE]: The skill instructions are limited to local file operations within a specifically configured vault directory. It reads from local markdown files (Context Index.md, project-context.md) and writes to a .canvas file (JSON format) located in the vault.
[SAFE]: No evidence of remote code execution, package installation, or network exfiltration was found. The skill does not request access to the network or use tools for external data transfer.
[SAFE]: No obfuscation techniques or prompt injection attempts were detected in the skill metadata or instructional content.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from the Obsidian vault (markdown notes) to extract wikilinks and note titles for graph generation.
Ingestion points: Reads note titles and wikilinks from vault notes and Indexes/Context Index.md.
Boundary markers: None present; the skill treats note content as source material for the graph.
Capability inventory: Limited to file reading and file writing of structured JSON data.
Sanitization: The output is constrained to a specific JSON schema (JSON Canvas), minimizing the risk of the data being interpreted as executable instructions by the agent or downstream tools.

vault-context-canvas