Implement a Flink streaming job that analyzes Google cluster trace data to find the longest session (stage) of task SUBMIT events per finished job. Use when the task involves Flink session windows on cluster trace task/job events, session gap detection, or counting events in session windows keyed by job ID.
Analyze USGS streamflow stations for flood events by comparing gage height data against NWS flood stage thresholds. Use when detecting flood days for USGS stations over a date range.
Generates a community pulse report for a GitHub repository by querying the GitHub Search API for PRs and issues within a date range, computing merge statistics, identifying top contributors, classifying bug reports by label substring, and counting resolved bugs. Use when asked to produce a JSON summary of repository…
Run and calibrate the General Lake Model (GLM) for lake temperature simulation. Use when task requires running GLM, computing RMSE against observations, and calibrating parameters to meet an RMSE threshold.
Gravitational wave detection using matched filtering with PyCBC. Loads .gwf data, conditions it, performs grid search over mass parameters for multiple approximants, and reports peak SNR per approximant.
Solve DC Optimal Power Flow with spinning reserves from MATPOWER-format network JSON. Produces economic dispatch report with generator outputs, reserves, line loadings, and operating margin.
HVAC PID temperature controller skill. Runs calibration, estimates first-order system parameters, tunes PID gains via IMC, executes closed-loop control, and computes performance metrics. Use for thermal control tasks with first-order plant dynamics.
Invoice fraud detection skill: extracts invoices from PDF, matches against vendor master and PO register, flags fraud using priority-ordered checks. Auto-calibrates fuzzy threshold and filters non-standard PO entries.
Extract dates and total amounts from scanned receipt images using OCR and write results to Excel. Use when processing receipt images for data extraction.
Harmonize clinical lab data with mixed units to US conventional units. Detects bimodal distributions from mixed unit systems, applies molecular-weight-based conversions, handles comma decimals and scientific notation.
Analyze lake surface water temperature trends and attribute warming to driver categories using Mann-Kendall trend test with year-based Sen slope, and per-category R-squared importance. All mappings derived at runtime.
Extract display-mode LaTeX formulas from PDF research papers. Use when the task requires identifying all display-mode (own-line) formulas from a PDF, extracting them as LaTeX, cleaning tags/punctuation, detecting bracket mismatches and misspelled commands, and outputting original + fixed formulas in markdown format.
Completes Lean 4 induction proofs for recursive sequence bounds. Use when a task requires proving an inequality about a recursively defined sequence (e.g., geometric series partial sums) in Lean 4 with the Math2001/Macbeth library tactics (simpleinduction, addarith, numbers, extra). Provides end-to-end proof…
Normalize noisy bilingual manufacturing defect reason texts into standardized codebook entries. Derives all categories, keywords, and matching rules from supplied codebook files at runtime. Policy parameters are configurable.
Analyze reflow oven thermocouple data, MES logs, and test defects to answer questions about preheat ramp rates, TAL, peak temperature, conveyor speed, and best run selection per the handbook specifications.
Repair FJSP baseline schedules to eliminate downtime violations while respecting freeze, right-shift, and policy budget constraints. Use when given manufacturing scheduling tasks with instance.txt, downtime.csv, policy.json, and baselinesolution.json.
Extract keyframes from a Super Mario video, convert to grayscale, count coins/enemies/turtles using template matching with NMS and SQDIFF filtering, and output CSV results. Use when given a video file with template images for object counting.
DBSCAN hyperparameter grid search for Mars cloud clustering with Pareto frontier extraction. Use when optimizing DBSCAN parameters (minsamples, epsilon, shapeweight) to cluster citizen science annotations against expert labels.
Multilingual video dubbing pipeline: TTS synthesis with Kokoro, temporal alignment to SRT windows, ITU-R BS.1770-4 loudness normalization, and video muxing. Use when tasked with dubbing a video from one language to another.
Fill OOXML (.docx) templates by replacing {{PLACEHOLDER}} tokens with data from a JSON file. Handles placeholders split across XML runs, conditional sections (IF/ENDIF blocks), and processes all document parts including headers and footers.
Organize heterogeneous documents (PDF, DOCX, PPTX) into subject-based folders using keyword-based text classification. Use when you need to sort documents into predefined categories based on their content.
Parallelize a sequential TF-IDF search engine using Python multiprocessing. Covers parallel index building (tokenization, TF, DF, IDF, inverted index, doc vectors) and parallel batch query search. Produces identical results to sequential with 1.5x+ indexing and 2x+ search speedup on 4 workers.
At most 3 mods per repository are shown here, and a mod shipped inside a plugin is left to that plugin's page — the rest are on their repository pages: