WorkflowTrim v1.0

WorkflowTrim v1.0: the version used for the results reported in the paper "WorkflowTrim: Automated Detection, Repair and Savings Measurement of Resource-Waste Smells in GitHub Actions CI/CD Workflows".

Tool

• Detector for eight resource-waste smells in GitHub Actions workflows (S1–S8; rules v0.4)
• Automated repair of six smells (S1, S2, S3, S5, S6, S8) with minimal, comment-preserving edits and four safety checks (parses, smell removed, steps unchanged, no new actionlint error)
• Savings estimate for S1 and S2 from a repository's real run history (minutes and list-price cost)
• Command line (python -m workflowtrim owner/repo) and Streamlit web interface (streamlit run app.py)

Study data and scripts

• RQ1: 817 repositories, 8,945 workflow files, 28,246 findings (repos.csv, findings.csv)
• RQ2: detector accuracy against two annotators, precision 0.80, recall 0.99 (rq2/)
• RQ3: repair of 9,128 files and manual review of 120 repairs, 100 judged correct (rq3/)
• RQ4: savings in the run histories of 20 repositories, 953.7 job minutes, 1.12% (rq4/)
• Statistics: confidence intervals and strata tests (stats.py)

65 automated tests (python -m pytest -q).

Authors: Ujjwal Raj, Vaibhav Mishra, Rishav Raj · Guide: Dr. Arvind Kumar · School of Computer Applications and Technology, Galgotias University