soloflow

Quickly capture multiple ideas and tasks from a braindump session

Identify, resolve, and verify a bug. Spawns bug-investigator → executor → shadow-verifier in sequence.

Propose learnings from completed sprint(s) in three buckets (clean-ups, backlog tasks, CLAUDE.md improvements) plus optional SoloFlow self-improvement feedback (tester mode), then apply what the user approves. Batches multiple pending sprints into one merged proposal when --all or multi-select subset is used.

Walk through SoloFlow configuration and adjust settings interactively

Clean up .soloflow/ state cruft (orphan plans, ghost sprint entries, stale stuck files, mid-commit settle crashes, empty epics, malformed queue entries)

Extract a structured idea from raw input, with optional external research

Scaffold or repair the .soloflow/ state directory and run the setup wizard

Unattended backlog-drain mode — runs executor → verifier → code-reviewer loops on all ready tasks, logs stuck/human-needed, never prompts. Visual verify off; parallel by default.

Generates the three "shared-context" docs every SoloFlow agent expects to load

Turn an approved idea into execution-ready task plans

Deep-clean the codebase — remove dead code, trim CLAUDE.md files, and verify no regressions

Lightweight bugfix path — skip refinement, go straight to executor + verifier

Triage pending human-review items — resolve cruft, walk Decisions/Actions/Testing/Deferred Visual buckets, refine found issues into backlog tasks

Generate a phased roadmap from a project vision, with deep questioning, parallel research, and epic generation

Phase 3 of the SoloFlow pipeline. Creates a sprint from ready tasks in the backlog and runs the executor → verifier → code-reviewer loop until the sprint is complete.

Show current SoloFlow sprint state and task progress

Re-sync MCP-dependent agent shadows in .claude/agents/ against the current plugin version

Check for SoloFlow updates and show the right command to run

Run standalone visual verification using Maestro MCP (mobile, falls back to CLI) and/or Playwright MCP (web)

Investigates a bug report — performs read-only root-cause analysis and produces a structured fix plan. Does NOT modify code.

Reviews CLAUDE.md improvement proposals against the existing codebase and CLAUDE.md files, producing tightly scoped diffs at the lowest appropriate directory level

Reviews all CLAUDE.md files for focus, trims redundancy, and identifies content that should be moved to specialized reference files

Reviews completed code for quality, reuse, and security

Reviews a codebase for redundancy, inefficiency, orphaned code, and dead exports, then produces a structured pruning proposal

Writes or updates unit and component tests for a single completed task's changes

Decomposes an approved idea into a coarse task skeleton (DAG, epic decisions, file-ownership hints) for downstream parallel detailers

Refines approved ideas into execution-ready plans with acceptance criteria, file ownership, and dependency mapping

Skeptical second-pass reviewer of a compound proposal; emits per-item IMPLEMENT / DONT_IMPLEMENT verdicts with concrete evidence. Read-only except for the proposal file where it inserts verdict blocks.

Analyzes completed sprints and out-of-scope findings, then produces a three-bucket proposal (clean-ups, backlog tasks, CLAUDE.md / CODE-PATTERNS.md improvements) for the user to approve

Implements a single task from an execution-ready plan with scope boundaries, deviation rules, and atomic commits

Extracts and structures raw ideas, feature requests, and bug reports into actionable task specs

Runs end-of-sprint integration and E2E tests, classifies failures as regressions or pre-existing

Synthesizes a roadmap brief and research into a phased set of epics with dependencies, milestones, and success signals

Gathers sprint close context and executes sprint close (mark complete, archive, commit, merge run branch) for the executor orchestrator

End-of-sprint aggregate code reviewer; surfaces cross-task quality and security findings as queued findings for the compounder

Gathers sprint context and executes sprint setup (branch, sprint.json, smoke tests) for the executor orchestrator

This skill should be used when a raw user idea is ambiguous and needs conversational disambiguation before structured extraction or planning. Provides a clarification loop with ambiguity checklist, scope-decomposition gate, one-question-at-a-time probing, and a readiness gate.

Deep conversational questioning to build a project roadmap brief before research and epic generation. Provides an 8-point checklist, scope-decomposition gate, one-question-at-a-time probing, and a readiness gate.

This skill should be used when verifying UI changes visually. Prefers Maestro MCP when available, falls back to the Maestro CLI. Web verification uses Playwright MCP. Provides patterns for availability probes, path selection, ephemeral/inline flows, screenshot capture, and view hierarchy inspection.

6 hooks across 5 events