The Agent Skills Directory

[DATA_EXFILTRATION]: The skill accesses session transcripts via the CLAUDE_TRANSCRIPT_PATH environment variable. These transcripts contain the full history of user interactions, which often includes sensitive data, source code, and configuration details shared during the session.
[PROMPT_INJECTION]: The skill creates an indirect prompt injection surface by reading untrusted session history and instructing the agent to evaluate it for patterns. * Ingestion points: evaluate-session.sh reads session history from the transcript file. * Boundary markers: Absent. There are no delimiters or instructions to the agent to ignore potentially malicious content within the transcript. * Capability inventory: The agent has filesystem access and shell execution capabilities. * Sanitization: Absent. The transcript data is not sanitized or filtered before being presented to the agent for analysis.
[COMMAND_EXECUTION]: The skill includes a shell script (evaluate-session.sh) designed to run as a platform hook. It performs local filesystem operations such as creating directories with mkdir and reading file contents with grep.

cc-skill-continuous-learning