菜单
Use when operating SkillX full-loop experiments after an inner-loop run completes, especially to audit result validity, classify failures, decide required/recommended reruns, prepare targeted rerun manifests, or determine whether it is safe to proceed to outer-loop optimization.
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
An overview of the python package for running the GaMMA earthquake phase association algorithm. The algorithm expects phase picks data and station data as input and produces (through unsupervised clustering) earthquake events with source information like earthquake location, origin time and magnitude. The skill explains commonly used functions and the expected input/output format.
An overview of the python package for running the GaMMA earthquake phase association algorithm. The algorithm expects phase picks data and station data as input and produces (through unsupervised clustering) earthquake events with source information like earthquake location, origin time and magnitude. The skill explains commonly used functions and the expected input/output format.
An overview of the core model API of SeisBench, a Python framework for training and applying machine learning algorithms to seismic data. It is useful for annotating waveforms using pretrained SOTA ML models, for tasks like phase picking, earthquake detection, waveform denoising and depth estimation. For any waveform, you can manipulate it into an obspy stream object and it will work seamlessly with seisbench models.
This is a summary the advantages and disadvantages of earthquake event detection and phase picking methods, shared by leading seismology researchers at the 2025 Earthquake Catalog Workshop. Use it when you have a seismic phase picking task at hand.
Clean messy tabular datasets with deduplication, missing value imputation, outlier handling, and text processing. Use when dealing with dirty data that has duplicates, nulls, or inconsistent formatting.
Difference-in-Differences causal analysis to identify demographic drivers of behavioral changes with p-value significance testing. Use for event effects, A/B testing, or policy evaluation.
| name | skillx-full-loop-audit |
| description | Use when operating SkillX full-loop experiments after an inner-loop run completes, especially to audit result validity, classify failures, decide required/recommended reruns, prepare targeted rerun manifests, or determine whether it is safe to proceed to outer-loop optimization. |
Use this skill after a SkillX inner-loop launcher finishes and before any outer-loop assignment or schema rewrite consumes the results.
Run the deterministic audit script first. Do not manually classify pair validity from dashboard impressions alone.
python3 scripts/audit_skillx_inner_loop_results.py \
--materialized-root <materialized-root> \
--run-label <inner-run-label>
The generic outer-loop wrapper already runs this audit by default:
scripts/run_skillx_outer_loop_step.sh <inner-run-label> <outer-label> <next-materialized-root>
By default, the wrapper stops before outer-loop work if rerun_required > 0.
Look under:
<materialized-root>/reports/<inner-run-label>/inner_loop_audit/
Key files:
inner_loop_audit.md: human-readable classification and rerun rationale.inner_loop_audit.json: source-of-truth structured audit.required_pair_manifest.json: pair ids that must be rerun.recommended_pair_manifest.json: pair ids that should be rerun if cost allows.run_required_rerun.sh: tmux launcher for required reruns.run_all_flagged_rerun.sh: tmux launcher for required plus recommended reruns.required: do not feed this result into outer-loop unless the user explicitly
overrides. Rerun first.recommended: result may contain useful signal, but a clean rerun is preferred.none: no rerun needed by default.Required examples:
Recommended example:
Non-rerun notes:
skipped_baseline_perfect is an intentional guard outcome.For the minimum repair set:
bash <audit-dir>/run_required_rerun.sh
For both required and recommended pairs:
bash <audit-dir>/run_all_flagged_rerun.sh
After the rerun finishes, rerun the audit on the effective run output before starting the outer-loop step.
For the observed round1 examples that define the current policy, see:
docs/plans/skillx/full-loop-post-inner-audit-reference-v0.1.md