Curated Claude Code skills with Korean and English variants. Includes dream (memory consolidation pass), session-seal (resumable session snapshots), session-retro (gated session retrospective that publishes reusable rules), to-html (gated HTML visualization of content), and task-routine (checklist-to-commit d.
A memory-maintenance workflow for Claude Code projects and users. It reviews recent conversation transcripts and existing memory files, combines useful information, removes outdated or conflicting entries, and updates the memory index.
Memory consolidation pass — a reflective reorganize of /.claude/memory/ files. Reviews recent transcripts and existing memory, merges new signals into topic files, prunes contradicted or stale entries, and tightens the MEMORY.md index. Args: (none) = current project memory; user = user-scope memory at…
A session review tool that examines what went well, what went wrong, what was difficult, and what was learned from both engineering and business viewpoints. It can suggest durable rules or decision records, but a person must approve anything before it is published.
Session retrospective — first scores whether previously published rules actually got reused, then analyses the session across 4 axes (went well / went wrong / was hard / learned) x 2 lenses (engineering / business), and finally puts promotion candidates in front of a human GATE before publishing anything as a rule…
A session-snapshot workflow that saves the original Claude Code conversation and records its purpose and approach in metadata. It can connect the snapshot to the most recent related snapshot in the same project.
Session snapshot sealer — captures the current Claude Code session's raw transcript and adds Why/How frontmatter (intent + approach) so a future session can resume cleanly. Saves under /.claude/sessions/. Proposes chain links (continues-why, continues-how) by scanning recent sealed snapshots. Use when the user says…
A development workflow that takes a task from design through testing, implementation, verification, review, and a Git commit. It can read tasks from a local Markdown checklist or a GitHub issue, and records recommendations, choices, and outcomes in an append-only log.
Drives a checklist of tasks to committed code, one task at a time — design, test strategy, implement, verify, review, commit — with an autonomy dial (manual / assisted / auto) deciding which gates run unattended, and an append-only decision log recording recommendation vs choice vs outcome. The task source is…
A tool that turns files or text into one readable HTML page when visual layout adds value. It compares options, provides adjustable controls, or combines diagrams, code, and warnings for learning.
Convert content (files, last output, pasted text) into a self-contained, human-readable HTML visualization — but only when it's worth it. First classifies the content against three patterns (Compare options side-by-side / Tune values interactively / Learn a concept with diagrams + code + caveats), estimates token cost…