The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze Claude Code conversation history logs stored in JSONL format. These files contain untrusted content from previous user interactions, subagent messages, and tool outputs which are treated as data but could be misinterpreted by the agent during analysis.
Ingestion points: The agent is instructed to read local history files located in ~/.claude/projects/ (e.g., in SKILL.md).
Boundary markers: The instructions provide jq and grep patterns to extract content but do not include explicit delimiters or safety instructions to prevent the agent from following commands embedded within the history being analyzed.
Capability inventory: The skill utilizes bash, jq, sed, grep, ls, and find to traverse the filesystem and process file contents.
Sanitization: There is no evidence of sanitization, escaping, or filtering of the external history content before it is presented to the agent for analysis.

cc-history