The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it processes web clippings, which are untrusted external data sources.
Ingestion points: Untrusted data enters the context in SKILL.md via the command obsidian read path="Clippings/<file>.md" during the processing workflow.
Boundary markers: The instructions do not define clear delimiters or provide instructions to the agent to disregard any commands or directives embedded within the clipping text.
Capability inventory: The skill uses the Bash tool to perform file system operations, including obsidian append, obsidian move, and obsidian delete as defined in SKILL.md steps 2 and 3.
Sanitization: There is no evidence of sanitization, filtering, or validation of the content read from the clippings before it is processed or integrated into the destination files.

process-clippings