The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from agent worktrees that could contain instructions designed to influence the ranking process or the LLM judge.
Ingestion points: Ingests data from git diff and agent-provided result files located at .agenthub/board/results/agent-{i}-result.md as described in SKILL.md.
Boundary markers: There are no explicit delimiters or instructions provided to the LLM to ignore potentially malicious embedded content within the results.
Capability inventory: The skill executes subprocesses via scripts/result_ranker.py and scripts/session_manager.py (referenced in SKILL.md).
Sanitization: No evidence of sanitization or validation is present for the ingested agent data before it is presented to the judge.
[COMMAND_EXECUTION]: The skill interpolates a configurable {eval_cmd} into a bash command executed via result_ranker.py. This represents a dynamic command execution surface where the specific command is determined by session-level configuration.

eval