The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted historical conversation data which could contain malicious instructions from previous sessions, potentially leading to an injection when the agent reads the resulting analysis reports.
Ingestion points: The scripts/analyze_history.py script reads the ~/.claude/history.jsonl file to parse past user interactions.
Boundary markers: The skill does not employ explicit boundary markers or instructions to the agent to ignore potentially malicious commands found within the historical data during processing or report generation.
Capability inventory: The skill utilizes the bash tool to execute the Python analysis script and the write tool to create report files on the local file system.
Sanitization: Historical text from the display field of the conversation history (which includes user prompts) is interpolated directly into the final conversation_analysis.txt report and printed to the terminal without escaping or sanitization.

conversation-analyzer