The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local Python scripts (result_ranker.py and session_manager.py) to manage session state and perform ranking logic.
[DYNAMIC_EXECUTION]: The skill supports a configurable evaluation command (eval_cmd) which is passed as an argument to a Python script and executed within the agent's worktree.
[INDIRECT_PROMPT_INJECTION]: The skill has a vulnerability surface for indirect prompt injection during the 'LLM Judge Mode'.
Ingestion points: The skill reads external data from git diff outputs and agent-generated markdown files located in .agenthub/board/results/.
Boundary markers: No specific boundary markers or sanitization logic are described to separate untrusted agent content from the judging instructions.
Capability inventory: The skill possesses shell execution capabilities via result_ranker.py and session_manager.py.
Sanitization: There is no evidence of sanitization for the content ingested from the agent worktrees before it is processed by the LLM judge.

eval