Reference

Keyboard Shortcuts

Global

Within Tabs

Tabs

Agent Defaults Operations

Configuration Files

Profile Schema

Execution profiles are YAML files with the following keys. All keys are optional – omitted keys cause the corresponding question to be asked interactively.

Identity

Task Selection

Branch & Worktree

Planning

Feedback

Exploration

QA

Manual Verification

Lock Management

Remote Workflow

These keys control behavior for the fully autonomous remote workflow (/aitask-pickrem):

Project Config Keys

Model Entry Schema

Each model in models_<agent>.json:

{
  "name": "opus4_7_1m",
  "cli_id": "claude-opus-4-7[1m]",
  "notes": "Most intelligent model for agents and coding",
  "verified": {
    "pick": 80,
    "explain": 80,
    "batch-review": 0
  },
  "verifiedstats": {
    "pick": {
      "all_time": {"runs": 5, "score_sum": 400},
      "month": {"period": "2026-03", "runs": 2, "score_sum": 180},
      "week": {"period": "2026-W11", "runs": 1, "score_sum": 100}
    }
  }
}

Verified Stats Buckets

See Verified Scores for a user-facing overview of how these scores are collected and what they mean.

Each skill entry in verifiedstats contains three time-windowed buckets:

Each bucket contains runs (number of ratings) and score_sum (sum of mapped scores, where raw 1-5 maps to 20-100).

The verified.<skill> field is always the rounded average of all_time and is maintained automatically for backward compatibility.

Old flat-format stats ({"runs": N, "score_sum": S}) are migrated automatically to the bucketed format on the next update.

All-Providers Aggregation

See Verified Scores for a user-facing explanation of provider-specific versus all-providers views.

To compute cross-provider stats for the same underlying LLM model:

1. Extract the model portion from each entry’s cli_id by stripping the provider/ prefix (e.g., openai/gpt-5.4 and opencode/gpt-5.4 both normalize to gpt-5.4)
2. Group entries with identical normalized model IDs across all models_*.json files
3. For each skill, sum runs and score_sum from the matching bucket across the group
4. For month and week buckets, only aggregate entries with the same period value

This aggregation is performed at read time by consumers — no duplicate aggregate values are stored.

ait settings implements this aggregation in the Models tab (cross-provider summary lines), the Agent Defaults tab (all-providers hints), and the model picker (Top Verified ranking).

Export Bundle Format

Config bundles use the .aitcfg.json extension and contain:

{
  "_export_meta": {
    "version": 1,
    "timestamp": "2026-03-03T12:00:00",
    "file_count": 5
  },
  "codeagent_config.json": { ... },
  "models_claudecode.json": { ... }
}

Bundles include only the files matching the default patterns: *_config.json, *_config.local.json, models_*.json, models_*.local.json.

Next: Workflows — end-to-end guides for using the TUIs together.