recursive-dep-trace

What problem does it solve?

This Skill produces a line-precise, reproducibility-focused audit of everything an entry script touches so a developer can determine exactly what is required to run that script from an empty machine. It replaces guesswork with a live, resumable markdown trace that proves which files, data assets, and third-party libraries are needed and where unresolved gaps remain.

Core Features & Use Cases

• Live, incremental trace: writes a stream-of-consciousness markdown file that is updated after every small read step so another agent can resume at the exact file:line and TODO stack state.
• Recursive static closure: reads every file to EOF and pushes every referenced file/path onto a TODO stack until the closure ends in third-party packages, public resources, or provable BLOCKED items.
• Data-file inspection: opens and inspects bundled data files (.npz, .pkl, .json, .yaml, .csv) to extract embedded path strings and continue tracing them.
• Robust BLOCKED handling: applies three distinct exhaustive search strategies before marking a reference blocked, logs commands verbatim, and performs a second-pass retry before user escalation.
• Use cases: preparing a repository for open-source release, extracting minimal reproducible examples, auditing training/inference pipelines, and verifying that a project can be run from a cold machine.

Quick Start

Trace recursive dependencies for train.py and produce a live incremental markdown trace at artifacts/task_recursive-dep-trace/train_trace.zh.md.