loop-audit 只给你 10 分。--tool grok--tool grok 脚手架未带 verifier,先停在 L2 assisted-fix。# Loop State — My Project Last run: 2026-07-16T09:00:00Z — 4 items surfaced, 2 promoted, 1 auto-fixed via minimal-fix, verifier approved. ## High Priority (loop 正在处理 / 等人) - [ ] flaky test tests/api/orders.spec.ts — 3 次失败,验证器要求人复核 - [ ] docs 链接失效 docs/getting-started.md#deploy — 已排队自动修 ## Watch List - lint warnings(从 Recent Noise 提升) ## Recent Noise (本轮忽略) - dependabot: 3 个 minor PR ## Post-Run Critique (上一轮复盘) - Adjustment: 把 infra check 加进扫描 - False positive: 1 CI flake(已知)
# Loop Configuration — Daily Triage ## Active Loops | Pattern | Cadence | Autonomy | Trigger | |--------------|---------|---------------|----------------| | daily-triage | 1d | L1 → L2 → L3 | cron 0 9 * * * | ## Human Gates - L1 report-only;L2 需 verifier 通过;L3 gate.yaml 内自动 - 高风险路径 → 人复核 ## Budget - 每日 tokens ≤ 100k(见 loop-budget.md) - 每轮追加 loop-run-log.md,kill switch: loop-pause-all ## Worktree isolation 每次 fix 走 loop-worktree create --run-id … --pattern daily-triage, 拒绝的尝试自动回收。
denylist: - secrets/ - .env - node_modules/ - prod/ - migrations/ auto_merge_allowlist: - docs/** - CHANGELOG.md - tests/fixtures/** budget: daily_token_cap: 100000 # from loop-cost "Suggested daily cap: 100k tokens" realistic_daily_tokens: 23000
loop-cost · daily-triage · L1 · 1d
| Early-exit / no-op | 5k / run |
| Full triage | 50k / run |
| Action every run | 200k / run * |
| Realistic blend | 23k / day |
| Suggested daily cap | 100k tokens |
STATE.md(当前状态)、LOOP.md(意图与 autonomy)、gate.yaml(机械护栏与预算),加一枚从 L0 10/100 一路爬到 L3 100/100 的 Loop Ready 徽章,直接贴进你的 README.md。