The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes existing notes to infer schemas through the schema_infer tool and validate conformance via schema_validate. While this involves ingesting potentially untrusted data from the knowledge graph, the actions are restricted to note management tools.
Ingestion points: Notes are read from the knowledge graph in SKILL.md (via schema_infer, schema_validate, and schema_diff).
Boundary markers: No explicit delimiters are specified for the processed content.
Capability inventory: Subprocess calls or eval/exec are not used; capabilities are limited to write_note, edit_note, and search_notes as documented in SKILL.md.
Sanitization: Relies on the underlying tool implementations for note handling.
[SAFE]: No other threat categories were triggered. The skill uses domain-specific tools for its stated purpose of schema management, and no external network operations or obfuscated content were detected.

memory-schema