AGENT SKILL

reflow-profile-compliance-toolkit

Deterministic handbook-grounded retrieval and thermocouple computations for ramp, TAL, peak, conveyor feasibility, and best-run selection; includes tooling/IO guardrails so outputs get generated.

License 许可
Apache-2.0
最近更新
2026-11-14
Stars
★ 0
Forks
1

01 前言 Front Matter

name
reflow-profile-compliance-toolkit
description
Deterministic handbook-grounded retrieval and thermocouple computations for ramp, TAL, peak, conveyor feasibility, and best-run selection; includes tooling/IO guardrails so outputs get generated.

03 SKILL.md 正文

Steps

  1. Tool/terminal execution discipline (mandatory):
    • If the terminal tool allows only one command per call, either (a) run one command per call, or (b) chain sub-commands into a single shell line using && or ;.
    • Prefer robust one-liners (example pattern): cmd1 && cmd2 && cmd3.
  2. Enumerate and validate inputs:
    • ls -lah /app/data
    • preview schemas with python -c or head to confirm required columns exist (especially thermocouples.csv: run_id, tc_id, time_s, temp_c).
  3. Extract handbook requirements into cfg (do not compute until this is filled from the handbook text):
    cfg = {
      "preheat_region": {...},
      "ramp_limit_c_per_s": ...,
      "tal_threshold_c_source": "solder_liquidus_c",
      "tal_min_s": ...,
      "tal_max_s": ...,
      "peak_margin_c": ...,
      "conveyor_rule": {...},
    }
    
  4. Prepare data deterministically:
    • cast run_id, tc_id to string;
    • sort thermocouples.csv by run_id, tc_id, time_s (stable mergesort).
  5. Compute metrics using the canonical recipes:
    • ramp: finite difference s_i = (T_i - T_{i-1})/(t_i - t_{i-1}), max over the handbook-defined preheat region (temperature/zone/time band as specified);
    • TAL: time_above_threshold_s(df_tc, liquidus_c) using linear interpolation and the handbook’s TAL window definition;
    • peak: per-TC max(temp_c), then run-level “coldest”/representative selection per handbook; if a run has no TC data, mark failing.
  6. Build outputs with JSON guardrails:
    • sort arrays/dict keys by ID ascending;
    • round to 2 decimals;
    • replace missing/invalid numeric results with null;
    • never claim compliance when the metric is null unless the handbook explicitly allows it.
  7. Write, then verify: create /app/output if needed, write each required JSON file, and confirm existence and valid JSON parse before responding.

Expected Result

A single coherent workflow that (1) successfully extracts handbook limits/definitions, (2) computes ramp/TAL/peak/speed feasibility deterministically from thermocouple traces, and (3) writes and verifies /app/output/q01.json–/app/output/q05.json without terminal/tool-call formatting failures or premature termination.

04 同仓库更多 Skills