The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill reads from historical interaction logs that could potentially contain malicious instructions or adversarial content from previous sessions.
Ingestion points: The skill reads from context/interactions/index.json and subsequent daily log files located at paths specified in the index.
Boundary markers: No explicit boundary markers or 'ignore' instructions are used when interpolating retrieved log content into the agent's response context.
Capability inventory: The skill performs local file read operations only; no network access or shell command execution capabilities are utilized.
Sanitization: There is no evidence of filtering or sanitizing the retrieved conversation content before it is presented to the agent or user.

interaction-recall