usage

Once yf is installed and the skills are deployed (install), the skills run inside your agent harness (Claude Code by default). Here are the paths you'll use most.

Set up a project

yf skills install                # deploy the skills into ~/.claude/{skills,rules}
yf doctor                        # verify the toolchain + every install

yf doctor should report your yf version, bd/uv/git present, and each skill ok.

yf-plan — structured, beads-tracked planning

/yf-plan replaces native plan mode with a scoped, reviewable, resumable pipeline whose work is tracked in beads.

/yf-plan build a rate limiter for the API      # scope -> investigate -> draft -> review
/yf-plan continue                              # resume an open plan
/yf-plan execute <plan-id>                     # (new session) run it in a git worktree
/yf-plan status                                # progress + parked plans

A plan is drafted and red-team reviewed, then — after you approve — executed in an isolated git worktree in a fresh session, merged back, and re-validated against the merged tree before anything is pushed. See yf-plan.

yf-research — cited, resumable deep research

/yf-research compare managed vector databases for RAG
/yf-research coordinate                        # (new session) run the pipeline to completion
/yf-research status

/yf-research decomposes a topic into a DAG of focused subtasks (retrieve → triangulate → synthesize → critique → refine → package) and produces a structured, citation-backed report with source-credibility scoring. See yf-research.

The beads loop

Beads-backed skills track their work in bd. The canonical loop, if you drive it directly:

bd ready --json                  # what's unblocked
bd update <id> --claim --json    # take a bead
# ... do the work ...
bd close <id> --reason "done"    # close it; dependents unblock

You rarely run these by hand during a skill run — the skill's coordinator drives the loop — but bd ready / bd show <id> are handy for seeing where a plan or research project stands.

Markdown + diagram tooling

The beads-free utility and markdown skills are one-shot:

/yf-markdown-lint docs/report.md               # validate GFM: links, tables, no wiki-links
/yf-markdown-pdf docs/report.md                # render to PDF (pandoc + xelatex)
/yf-diagram-authoring                          # author a d2 diagram -> light-mode PNG

Where to go next

• skills — every skill, generated from its SKILL.md: what it does, when it fires, and when to skip it.
• lifecycle — the install → preflight → invoke → execute loop in detail.
• architecture — how the yf kernel, embedded skills, beads, and upstream tracking fit together.