菜单
Run the backtracking delayed-label intelligence loop. Use when new analyst labels are available or to rewind causal reasoning on past detections.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Install, unload, and check status of the indago macOS LaunchAgents (AIS stream + R2 sync). Use when setting up a new machine, after a plist change, or to verify streams are alive and uploading.
Download and inspect the latest regional watchlists on R2 (arktrace-public or maridb-public). Use when finding demo MMSIs, checking sanctions_distance and ownership_chain, or verifying data-publish output before arktrace smoke tests.
Fetch and interpret indago data-publish metrics (P@50 email, R2 snapshots, lead time). Use after data-publish CI, when explaining metric deltas, or before updating external docs.
Generate an OSINT intelligence report for a vessel or watchlist batch. Use when investigating a flagged vessel or preparing a weekly analyst report.
Run the indago 11-step data pipeline. Use when ingesting new AIS data, refreshing features, or running a full regional screening.
Run the indago test suite. Use when changing pipeline logic, scoring models, or schema contracts.
| name | indago-run-backtrack |
| description | Run the backtracking delayed-label intelligence loop. Use when new analyst labels are available or to rewind causal reasoning on past detections. |
| license | Apache-2.0 |
| compatibility | Requires Python >=3.12 and uv; scoring model must have run first |
| metadata | {"repo":"indago"} |
uv run python scripts/run_backtracking.py --days 30
Checks for new analyst labels, propagates them backward through the causal model, and flags precursor signals that should have fired.
uv run python scripts/check_score_regression.py
See references/backtracking-runbook.md for full delayed-label loop operations, evidence types, and demo scenario.