The Agent Skills Directory

[DATA_EXFILTRATION]: The skill reads conversation transcripts stored in ~/.claude/projects/. These files contain the full history of AI interactions, including code snippets, configuration, and any sensitive data discussed during previous sessions.\n- [COMMAND_EXECUTION]: The skill executes a local Python script to process log files and provides documentation for running it via SSH piping to aggregate logs from remote systems.\n- [REMOTE_CODE_EXECUTION]: The documentation describes a pattern for executing the harvesting script on remote hosts via SSH piping (ssh <host> \"python3 -\" < harvest.py). While intended for legitimate administrative use, this involves executing local scripts on remote environments.\n- [PROMPT_INJECTION]: The skill processes untrusted historical session logs, creating a potential surface for indirect prompt injection.\n
Ingestion points: scripts/harvest.py reads all .jsonl files within the ~/.claude/projects/ directory.\n
Boundary markers: No explicit delimiters or instructions to ignore embedded commands are present in the ingestion logic.\n
Capability inventory: The skill has read access to the local file system and generates shell commands for the agent to execute.\n
Sanitization: User prompt snippets are filtered to exclude system blocks and are truncated to 70 characters.

mission-log