The Agent Skills Directory

[COMMAND_EXECUTION]: The skill documentation provides shell command templates using standard Unix utilities like jq, rg, awk, and grep to search and analyze session logs. These operations are restricted to the agent's internal session directory and are used for read-only analysis.
[PROMPT_INJECTION]: The skill is subject to indirect prompt injection because it analyzes historical conversation logs in ~/.otto/agents//sessions/*.jsonl which may contain instructions from past user messages. * Ingestion points: JSONL session log files in the agent's data directory. * Boundary markers: None present in the command templates to differentiate historical instructions from current ones. * Capability inventory: Reading and processing files via shell utilities. * Sanitization: Employs jq filtering to isolate text content, providing basic structural validation without content sanitization.

session-logs