MoJi

What Problem It Solves

Daily engineering knowledge often stays fragmented across chat logs, browser tabs, git commits, and verbal explanations. It is easy to lose after a few days. When reviewing work later, git log shows what changed but not why; when explaining a feature to a teammate, it is easy to fall back into long, repetitive explanations. This skill addresses the problem of fragmented, hard-to-search, hard-to-share, and hard-to-reuse work experience: it turns local work activity into reviewable and shareable text.

How It Works

It is a lightweight personal experience collector that runs against local project directories. Its core capabilities include:

• File change monitoring: scans configured working directories and uses file hashes to detect added, modified, and deleted files, not just commit messages.
• Git change records: collects commits and uncommitted changes from the last 24 hours, adding code-level context to notes.
• Shell history collection: records recently executed commands, preserving manual operations and ad-hoc debugging traces.
• Learning record collection: gathers experience entries in self-improving-agent format, giving reusable knowledge a consistent structure.
• Structured output: generates Markdown notes by date and stores them by category, making later retrieval and export easier.

The typical workflow is to add project directories to the watch list, run once to establish a baseline, and then schedule periodic collection. The output is usually day-level Markdown files that can be opened in Obsidian as a Vault, or exported for teammates and future project reviews.

Boundaries and Caveats

It is suitable for personal knowledge accumulation, project retrospectives, and team knowledge transfer, but it is not a full project management system. It depends on local directory access and shell history, so sensitive paths, private credentials, and production configuration should not be included in collection scope. Generated notes are closer to working drafts than finished design documents; before sharing, they still need business context, trade-off reasoning, and clear conclusions.