connorpham
UserOne command puts a whole virtual AI team (PM·BA·SA·DEV·QA) to work on your repo — with machine gates that can actually go red, evidence that survives the session, and autonomy with a paper trail. Works with Claude Code, Cursor, Windsurf, Codex, Copilot.
Categories
Indexed Skills (10)
ba
BA pipeline — turns the upstream spec into a runnable backlog that /dev and /qa can execute. Reads the spec + schema + existing backlog → shards the spec into docs/specs/<feature>.md (index + per-feature files, the repo's citable oracle) → decomposes a feature into user stories with TESTABLE acceptance criteria (Given/When/Then, each tracing to a spec section) → flags every gap/contradiction as a question for the owner instead of inventing requirements → challenger review → creates the tickets per the B4 gate (owner approves the draft when present; in owner-absent sessions at full autonomy, challenger APPROVE suffices; dry-run always stops at the draft).
dev
Ticket-driven DEV pipeline, end-to-end — preflight (tracker/design-source/git/DB pinged for real) → fetch the ticket → read spec + schema (the ticket is a claim, the spec is the spec) → UI tickets pull the design from the ticket's design link (design source = visual oracle; exact colors/spacing/text from node data) → implement on a feature branch → /verify gate + evidence → self-review with machine-measured design fidelity → 2 fresh review agents must approve (+1 on high-stakes diffs; committed review trail enforced at pre-push) → push + PR → mandatory 7-part plain-language report comment on the ticket, then transition to In Review (Done belongs to /qa). Runs the machine DoR gate and claims the ticket before coding.
docs
Documentation bootstrapper for mature-but-undocumented repos — the 'read then ask' lane. Reads the codebase FIRST (structure, manifests, routes/entrypoints/models/migrations, existing README, git-log themes) and drafts a 'what this system appears to do' map where every inference is marked ⚠ UNVERIFIED → interviews the owner in ONE batched question table (each question carries what the code suggests, what is ambiguous, and a proposed answer acceptable in one word) → writes the standard oracle the other lanes assume exists: specs INDEX + per-area shards (DRAFT-FROM-CODE vs OWNER-CONFIRMED), decisions.md seeds, knowledge-base.md seeds, known-issues.md → proposes a vteam.config.yaml patch (high_stakes_terms/paths, specs.sources, docs.task_context) for the owner to apply. Writes documentation only — never product code, never the config itself.
plan
Greenfield planning intake — for projects that have NEITHER code NOR documentation. Interviews the owner section by section (a five-field kernel: Why, Capabilities, Constraints, Non-goals, Success signal), with a one-round elicitation menu after every drafted section → writes a BRIEF, then a PRD whose requirement rows carry gate-compatible codes, then an optional architecture spine → registers the PRD as a SOURCE document (specs.sources) so the verbatim gate guards everything /ba later shards from it. Writes planning documents only — never product code, never tickets, never the config itself.
pm
PM+SA orchestration lane — one command runs the whole virtual team so the owner only watches results. Reads the tracker + the sprint plan + the decision queue → picks the highest-value UNBLOCKED work → dispatches to the right lane (/ba for backlog, /dev for code, /qa for verification, the SA lane for ADRs) with every child gate intact → ends each session with the 'your desk' report: done / needs-you / upcoming deadlines. Blocked work stays blocked — open business questions are collected for the owner, never answered by the machine.
qa
VERIFY-ONLY QA pipeline for a ticket a dev claims done (the spec is the oracle). Reads the ticket + spec + schema to derive expected behavior → designs 2–5 test cases (exact repro, boundary, whole-screen sanity, read-only DB verify for writes) → self-provisions missing data through the REAL UI flow (write-gated) → runs them HEADED in the browser → collects evidence (named screenshots, annotated images with in-image captions) → cross-checks every ticket claim against an evidence file → machine gate (evd_check.py) → challenger sign-off → plain-language REPORT.md anyone can read → tracker comment posted once both machine gates are green, ticket transition per the full-auto policy. NO code change, NO fix — verification only.
team
One command puts the whole virtual team to work. /team runs a full 'workday' — clears the owner's decision queue first (batched questions), then works through every UNBLOCKED item (DEV tickets sequentially via /dev, BA drafts and SA ADRs in parallel background worktrees, QA verification between dev tasks) until everything left needs the owner, then prints one end-of-day desk report. The autonomous entry point on top of /pm — same rules, same gates, same decision queue.
verify
Create and run the verification gate for this repo — the profile's ordered step manifest (ledgers → lockfile → lint → types → unit → build → reality checks → integration → e2e), with exact closing lines recorded. Also the standard for WRITING tests — expected values cite the spec or the schema, every behavior gets a boundary pair, no tests that mirror the implementation. Invoked standalone and as the /dev verify step, and before declaring ANY code change done.
guidelines
Behavioral guidelines to reduce common LLM coding mistakes. Loaded by /dev before planning; applies when writing, reviewing, or refactoring code — avoid overcomplication, make surgical changes, surface assumptions, define verifiable success criteria.
setup
Bootstrap the vteam proof-of-done harness in the current repo — grade it first (npx vteam-harness audit), install (npx vteam-harness init), verify (npx vteam-harness doctor). Use when the user asks to set up vteam, add proof-of-done gates, or install the virtual AI team.
Bio shown is the top-scored skill's repo description as a fallback — real GitHub bios land in a future update.