Skip to content

anbeime/skill / skills/data-flow-skill/data-flow-skill

data-flow-skill

Verified

面向数据分析任务的一站式 Agent Skill,覆盖数据类型检测、数据理解、预处理、统计分析、可视化图表生成、正式报告和幻灯片输出。适用于课程作业、论文实验、业务分析、SEO/GEO 数据报告和可追溯数据分析工作流。

FileKindSize
README.mdmarkdown1.8 KB
SKILL.mdskill md7.7 KB
references/data-types.mdmarkdown907 B
references/reporting.mdmarkdown816 B
references/slides.mdmarkdown895 B
references/validation.mdmarkdown1.1 KB
references/visualization.mdmarkdown797 B
references/workflow.mdmarkdown875 B
requirements.txtmanifest97 B
scripts/README.mdmarkdown1.0 KB
scripts/analysis/__init__.pyscript35 B
scripts/analysis/analyzer_data_understanding.pyscript5.2 KB
scripts/analysis/analyzer_dataset_detection.pyscript7.6 KB
scripts/analysis/analyzer_findings_generation.pyscript57.1 KB
scripts/analysis/analyzer_preprocessing.pyscript1.9 KB
scripts/analysis/analyzer_statistical_analysis.pyscript934 B
scripts/analysis/analyzer_strategy_registry.pyscript3.4 KB
scripts/analysis/analyzer_time_series_analysis.pyscript2.0 KB
scripts/image_gen/image_generator.pyscript6.4 KB
scripts/mermaid/__init__.pyscript202 B
scripts/mermaid/cli.pyscript4.8 KB
scripts/mermaid/echarts_export.pyscript6.6 KB
scripts/mermaid/flowchart.pyscript9.4 KB
scripts/visualization/matplotlib/bar_memevolve.pyscript4.7 KB
scripts/visualization/matplotlib/bar_spice.pyscript6.0 KB
scripts/visualization/matplotlib/box_plot.pyscript4.4 KB
scripts/visualization/matplotlib/bubble_chart.pyscript5.8 KB
scripts/visualization/matplotlib/calendar_heatmap.pyscript8.2 KB
scripts/visualization/matplotlib/line_aime.pyscript3.0 KB
scripts/visualization/matplotlib/line_loss_inset.pyscript5.4 KB
scripts/visualization/matplotlib/line_selfdistill.pyscript7.0 KB
scripts/visualization/matplotlib/parallel_coordinates.pyscript6.7 KB
scripts/visualization/matplotlib/radar_dora.pyscript4.7 KB
scripts/visualization/matplotlib/scatter_break.pyscript5.8 KB
scripts/visualization/matplotlib/scatter_tsne.pyscript4.4 KB
scripts/visualization/matplotlib/stacked_bar.pyscript4.5 KB
scripts/visualization/matplotlib/violin_plot.pyscript5.3 KB

scripts/visualization/matplotlib/line_aime.py

View on GitHub
"""
Reproduce: image6.png — AIME avg@32 training curve
Two lines with vertical breakpoint markers + horizontal reference line.
Style: sans-serif, 4-spine box, no grid, right-bottom legend.
"""

import numpy as np
import matplotlib.pyplot as plt
import matplotlib.lines as mlines

plt.rcParams.update({
    'font.family': 'sans-serif',
    'font.sans-serif': ['DejaVu Sans', 'Arial', 'Helvetica'],
    'text.usetex': False,
})

rng = np.random.default_rng(42)

# ---- 模拟数据 ----
# w/ Dynamic Sampling (purple): 0-2200 steps, rises fast, ends ~0.42
steps_dyn_a = np.arange(0, 2300, 50)
y_dyn_a = 0.43 * (1 - np.exp(-steps_dyn_a / 600))
y_dyn_a += rng.normal(0, 0.012, len(steps_dyn_a))
y_dyn_a = np.clip(y_dyn_a, 0, 0.45)
y_dyn_a[:2] = [0.03, 0.03]   # cold start

# w/o Dynamic Sampling (cyan): 0-9000, slower rise, peaks ~0.42 at step 6000, then drops
steps_nodyn = np.arange(0, 9100, 100)
y_nodyn = 0.38 * (1 - np.exp(-steps_nodyn / 1200))
y_nodyn += rng.normal(0, 0.012, len(steps_nodyn))
y_nodyn = np.clip(y_nodyn, 0, 0.44)
y_nodyn[:2] = [0.01, 0.02]
# After step 6000, add gradual decline
mask = steps_nodyn > 6000
y_nodyn[mask] -= 0.06 * (steps_nodyn[mask] - 6000) / 3000

C_DYN   = '#5B0DAD'   # 更深紫,接近原图
C_NODYN = '#5BBCCA'   # 柔和青绿
C_REF   = '#3D78C2'   # 独立蓝色(参考线,与曲线色区分)
STEP_DYN  = 2200      # 紫色垂直线
STEP_NODYN = 6050     # 青色垂直线
REF_Y = 0.43          # 水平参考线

fig, ax = plt.subplots(figsize=(9.0, 4.8))

# ---- 两条主线 ----
ax.plot(steps_dyn_a, y_dyn_a, color=C_DYN, lw=1.4, zorder=3, label='w/ Dynamic Sampling')
ax.plot(steps_nodyn, y_nodyn, color=C_NODYN, lw=1.4, zorder=3, label='w/o Dynamic Sampling')

# ---- 水平参考线 ----
ax.axhline(REF_Y, color=C_REF, lw=1.5, linestyle='--', zorder=2)

# ---- 两条垂直虚线 ----
ax.axvline(STEP_DYN, color=C_DYN, lw=1.5, linestyle='--', alpha=0.85, zorder=2)
ax.axvline(STEP_NODYN, color=C_NODYN, lw=1.5, linestyle='--', alpha=0.85, zorder=2)

# ---- Axes 样式 ----
ax.set_xlim(-100, 9200)
ax.set_ylim(-0.01, 0.47)
ax.set_xticks([0, 2000, 4000, 6000, 8000])
ax.set_xticklabels(['0', '2000', '4000', '6000', '8000'], fontsize=10)
ax.set_yticks([0.0, 0.1, 0.2, 0.3, 0.4])
ax.tick_params(labelsize=10, direction='out', length=4, width=0.8)
ax.set_xlabel('Step', fontsize=12)
ax.set_ylabel('AIME avg@32', fontsize=12)

# 四边框(all spines visible)
for sp in ax.spines.values():
    sp.set_visible(True)
    sp.set_linewidth(1.0)

ax.grid(False)

# ---- 图例 ----
leg = ax.legend(
    loc='lower right',
    fontsize=9.5,
    frameon=True,
    facecolor='white',
    edgecolor='#AAAAAA',
    framealpha=1.0,
    borderpad=0.5,
    labelspacing=0.3,
    handlelength=2.0,
    handletextpad=0.5,
)

from pathlib import Path

output_path = Path('output/figures/line_aime_repro.png')
output_path.parent.mkdir(parents=True, exist_ok=True)
fig.tight_layout(pad=0.8)
fig.savefig(
    output_path,
    dpi=300, facecolor='white',
)
plt.close(fig)
print('saved: line_aime_repro.png')
reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis.Stars: 466kUpdated Claude CodeCodexVerifiedInstall
reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis.Stars: 466kUpdated Claude CodeCodexVerifiedInstall
diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness.Stars: 319kUpdated Claude CodeCodexVerifiedInstall
dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependenciesStars: 319kUpdated Claude CodeCodexVerifiedInstall