The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests data from user conversations and stores it in persistent files that are later read to influence agent behavior.
Ingestion points: Daily notes and conversation summaries are written to files in $AGENT_HOME/memory/ and $AGENT_HOME/life/.
Boundary markers: The skill employs a structured YAML schema for atomic facts in items.yaml, which provides structural delimitation.
Capability inventory: The skill utilizes file read/write operations within the agent's home directory and executes the qmd CLI utility.
Sanitization: The instructions do not specify explicit sanitization or filtering of user-supplied content before it is stored in the knowledge graph.
[COMMAND_EXECUTION]: The skill relies on the qmd command-line tool for semantic, keyword, and vector searches (e.g., qmd query, qmd search, qmd vsearch) and for indexing the personal directory (qmd index $AGENT_HOME).
[DATA_EXPOSURE]: As its primary function, the skill stores extensive personal context, including factual data, relationships, and user behavioral patterns (tacit knowledge), in plain text and YAML files on the local file system.

para-memory-files