sohei56
UserMulti-agent AI Scrum team for Claude Code — native Mac app + CLI. Autonomous Sprints; you stay Product Owner.
Categories
Indexed Skills (20)
cleanup-audit
Multi-agent parallel investigation that finds bugs, drift, dead code, redundant docs, and stale references in this repo. Spawns 8 read-only sub-agents on independent axes (stale-refs first, then consistency × 3, redundancy × 2, dead-artifact × 2), deduplicates findings, classifies into 5 tiers, and surfaces user-only decisions. Use after major refactors, before releases, or for periodic repo hygiene checks.
backlog-refinement
Refine coarse-grained PBIs into implementation-ready items
change-process
FR-016 Change Process — modify frozen documents with user approval
codebase-audit
Whole-repo, multi-agent audit that IS the Sprint-end cross-review ceremony (product-wide integrity): 4 axes — spec-conformance, logic/defect hunt, redundancy, and product-security — over the ACCUMULATED codebase at HEAD, not the Sprint diff. Findings are swept to zero per defect class — one class = one PBI covering every occurrence, documentation drift batched into a single DOCS PBI — and non-blocking: EVERY finding is PO-adjudicated, Critical/High carrying a next-Sprint recommendation, and a rejected finding is suppressed with a recorded decision. At Integration-Sprint entry a thin re-check confirms the latest audit is fresh and no open blocking (non-Low) audit PBIs remain before testing proceeds.
create-brief
Co-author a product brief (docs/product/brief.md) with the human through a structured interview, then quality-gate it against best practices. The brief is the single human-authored input that anchors every scope / YAGNI decision the team makes — in both human and autonomous modes. It anchors the Requirement Definition interview (the Requirements Analyst reads it first and reconciles requirements against it) and, in autonomous mode, is the Product Owner's source of truth. Use when no brief exists yet, when a brief is thin or unmeasurable, or when the operator runs `/create-brief`.
cross-review
Sprint-end product-wide integrity gate. The five per-aspect reviews (requirement conformance, functional quality, security, maintainability, docs consistency) run PER-PBI inside the pipeline before a PBI reaches awaiting_cross_review. Sprint-end cross-review is the whole-repo codebase-audit ONLY: static analysis + 4 audit axes (spec-conformance, logic-defect, redundancy, product-security) over the accumulated codebase at HEAD. The audit is non-blocking — every finding is PO-adjudicated and the ones routed to next_sprint become draft PBIs — except the documentation batch, which is closed inside this Sprint (Step 7b) so the drift does not compound while it waits; it never reverts a PBI.
install-subagents
Select and verify project-managed sub-agents for PBI work. Developers invoke after receiving PBI assignments.
integration-tests
Integration Tests — design-driven, systematic verification for the Integration Sprint. Derives boundary-value, flow-branch, and pattern-branch test cases from the enabled design specs, builds stubs for non-reproducible external interfaces, automates API and UI tests as committed project assets, and records results to .scrum/test-results.json. Runs before UAT & Release.
pbi-escalation-handler
Handles PBI pipeline escalation notifications from Developer. Reads escalation context, applies response matrix (retry / split / hold / human), and routes to user when human intervention is needed.
pbi-merge
SM-side merge orchestration for a single PBI. Triggered when the Developer notifies `[<pbi-id>] PBI_READY_TO_MERGE`. Drives `.scrum/scripts/merge-pbi.sh` and handles the failure / retry cycle through SendMessage to the assigned Developer.
pbi-pipeline
PBI development pipeline — orchestrates design, impl+UT, PBI review, and UT-run stages with sub-agent fan-out, file-based handoff, and deterministic termination gates (Anthropic + Ralph + GAN-derived). Used by Developer per assigned PBI. Replaces former design + implementation skills.
po-acceptance
PO acceptance verification — launches and operates the app to verify acceptance criteria by command execution. Used in Sprint Review (demo mode) and Integration Sprint (UAT mode) when .scrum/config.json po_mode is "agent".
requirement-definition
Requirement Definition ceremony. Spawns a single requirements-analyst to elicit requirements through natural-language dialogue and mandatory benchmark web-search research, producing a requirements document, a benchmark findings document, and the initial Product Backlog.
retrospective
Sprint Retrospective — record improvements for the next cycle
scaffold-design-spec
Create template stub files for newly enabled catalog entries, including user-facing documentation
smoke-test
Smoke Test — automated test execution for Integration Sprint. Detects test frameworks, runs all tests, performs HTTP smoke testing, optionally runs browser E2E via Playwright MCP, and records results to .scrum/test-results.json.
spawn-teammates
Reproducible teammate creation during Sprint Planning. Reads Sprint and Backlog state, spawns Developer teammates via Agent Teams with consistent naming and assignment.
sprint-planning
Sprint Planning ceremony — select PBIs, assign developers, create Sprint
uat-release
UAT & Release — user-acceptance testing of the integrated product followed by the release decision. Entered from Integration Tests once automated tests pass; drives UAT, defect collection, defect→PBI routing, and the go/no-go release gate.
sprint-review
Sprint Review ceremony — present Increment to user
Bio shown is the top-scored skill's repo description as a fallback — real GitHub bios land in a future update.