The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection attack surface because it is designed to ingest and process data from external websites and user-provided URLs.
Ingestion points: Web search results gathered in Step 1 and content from user-provided URLs used as triggers.
Boundary markers: The skill instructions do not specify the use of delimiters or 'ignore embedded instructions' warnings when processing fetched content.
Capability inventory: The skill utilizes search_notes, write_note, and edit_note to read from and write to the local knowledge base (SKILL.md).
Sanitization: There is no explicit requirement for the agent to sanitize, escape, or validate the content fetched from the web before summarizing it or storing it in a note.
[COMMAND_EXECUTION]: The skill invokes specific tools to perform file-based operations in the local environment for its intended purpose.
Evidence: Invocations of search_notes, write_note, and edit_note are used to manage the lifecycle of 'Basic Memory' entities across different directories like 'organizations', 'people', and 'concepts'.

memory-research