The Agent Skills Directory

Indirect Prompt Injection (SAFE): The skill processes user-controlled documentation (e.g., docs/AGENTS.md), which is an indirect prompt injection surface. Ingestion points: docs/AGENTS.md and other markdown files in the docs/ directory. Boundary markers: Absent; the agent reads the files directly for comparison. Capability inventory: The skill uses Task objects to perform file reads and directory listings. Sanitization: None; however, the output is a plan presented to the user for approval, mitigating the risk of autonomous malicious action.
Dependency Analysis (SAFE): The skill relies on a local sibling skill (skills/obsidian-plan-wiki/) for its canonical specification. No external or untrusted packages are downloaded.
Network Security (SAFE): No network operations, API calls, or data exfiltration patterns were found.
Privilege Management (SAFE): The skill operates within the standard file access scope required for its stated purpose without attempting privilege escalation.

obsidian-upgrade