The Agent Skills Directory

[COMMAND_EXECUTION]: Executes shell commands (find, wc, stat) and Node.js one-liners to collect and parse local session data. These operations are restricted to the user's home directory and are necessary for the skill's primary function.
[PROMPT_INJECTION]: Potential for indirect prompt injection as the skill processes historical session logs. Ingestion points: reads files from ~/.claude/projects/ containing past interactions. Boundary markers: data is extracted using structured JSON parsing before being summarized by the LLM. Capability inventory: limited to local file metadata and JSON parsing via Node.js. Sanitization: uses specific JSON field extraction to filter content before analysis.

session-insights