Skip to Content

Synopsis

aigon feature-eval <ID> [--agent=<id>] [--allow-same-model-judge] [--force]

Shortcuts

  • /afe (slash command) · aigon afe (CLI)

Flags

FlagDescription
--agent=<id>Force a specific evaluator agent instead of auto-selection
--allow-same-model-judgeAllow the evaluator to be the same model as an implementer
--forceRe-run evaluation even if a winner has already been recorded

Description

feature-eval is the Fleet mode judging step — it compares all agent implementations against the spec, picks a winner, and records that decision so feature-close knows which branch to merge. Run it from the main repo after all agents have signalled submitted.

By default, Aigon auto-selects an evaluator that is different from the implementing agents to avoid bias. Pass --agent=<id> to override and choose a specific evaluator. To allow the same model to judge its own implementation (not recommended), use --allow-same-model-judge.

In Drive mode (single agent), use feature-code-review instead — there is no competition to judge.

What it does

  1. Reads all submitted implementations from the feature’s worktrees
  2. Selects an evaluator (or uses the one specified via --agent)
  3. Opens an evaluation session pre-loaded with the full diff for each agent
  4. Evaluator compares implementations against the spec’s acceptance criteria
  5. Records the winner in workflow state
  6. Suggests feature-close <ID> <winning-agent> as the next step

Examples

# Auto-select an evaluator and compare aigon feature-eval 42 # Force a specific evaluator aigon feature-eval 42 --agent=gg # Re-evaluate after changing your mind about the winner aigon feature-eval 42 --force

After evaluation

# Merge the winner aigon feature-close 42 cc # Merge winner and adopt selected pieces from losers aigon feature-close 42 cc --adopt gg
Last updated on