By preql-data
Mandatory orchestrator -> specialists -> QA workflow with Beads task tracking
Pin a new model identifier across all workflow agents (orchestrator, qa, backend, frontend, devops, and the future grader). Claude-invokable; the user types intent in plain English and Claude calls this when a newer model lands.
Invoke the Phase C mutation harness against the active project.
Workflow orchestrator. Coordinates work and delegates to specialist subagents (@backend, @frontend, @devops, @qa); does not implement code directly. Use proactively as the first responder for any non-trivial software-engineering request — analyzing intent, opening Beads tasks, and routing the work.
Quality assurance specialist and the mandatory quality gate. Reviews and tests code changes, validating user-visible behavior rather than implementation details. Use proactively whenever code has been modified and needs validation before delivery, or whenever a Beads task is labelled `qa-pending`.
Backend engineering specialist. Implements server-side logic, APIs, databases, authentication, and background jobs, and updates Beads with structured progress notes. Use proactively whenever a request involves server-side concerns or a Beads task is labelled `backend`.
Frontend engineering specialist. Implements UI, components, styling, accessibility, and client-side state, and updates Beads with structured progress notes. Use proactively whenever a request involves user-facing interfaces or a Beads task is labelled `frontend`.
DevOps specialist. Handles infrastructure, CI/CD, Docker, deployment, hooks, and tooling, and updates Beads with structured progress notes. Use proactively whenever a request involves infrastructure, build, deploy, or hook concerns or a Beads task is labelled `devops`.
Admin access level
Server config contains admin-level keywords
Executes bash commands
Hook triggers when Bash tool is used
Own this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimOwn this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimBased on adoption, maintenance, documentation, and repository signals. Not a security audit or endorsement.
Modifies files
Hook triggers on file write and edit operations
Modifies files
Hook triggers on file write and edit operations
Uses power tools
Uses Bash, Write, or Edit tools
Uses power tools
Uses Bash, Write, or Edit tools
A plugin for Claude Code that turns "build me a feature" into a tracked, reviewed, regression-tested change set — without you driving every step.
qa-approved on the active task. The decision
is recorded in Beads as a label and a comment — full audit trail, no
honour system. The label alone is not enough: the Stop hook releases
only when the approval comment carries a change_set_hash matching the
current diff, so a forged qa-approved label (added without
qa-gate.sh approve) does not pass the gate.bd update --status in_progress. GitHub issue auto-linking is provided
(bd-github-link.sh: posts a back-link comment on task close and parses
Closes #N); its call shapes are verified by the bd-github-link L1
test against a stubbed gh, not a live round-trip. Pick up tomorrow
where you left off tonight.grader subagent scores the work against
a versioned rubric (default + per-domain overlays + a bugfix overlay
for evidence-before-fix). The grader sees only the diff, the SPEC,
and the lessons ledger — no specialist conversation context — so its
verdict is independent. Iteration cap is binding and engages the
escalation path on cap-hit.bd-mcp exposes 21 typed Beads
tools (no shell quoting bugs). code-graph-mcp exposes 7 graph tools
(code_search, code_context, symbol_callers, impact_of,
dead_code, dependency_path, code_index_health) backed by a
tree-sitter + SQLite index. Both load automatically via .mcp.json
and ${CLAUDE_PLUGIN_ROOT}./mutation-sweep generates fault-class mutants for hook scripts,
runs them in throwaway git worktrees against the free L1/L2 tiers,
and routes survivors through a read-only judge subagent
calibrated against a hand-labeled set (precision ≥ 0.8). Every
step is dev-cycle-manual — no CI wiring, no scheduled job, no
automatic paid call.LESSONS.md is
append-only via lessons.sh add; the orchestrator reads it before
decomposing non-trivial work; the grader reads it as part of every
grading packet. The two seed lessons (parallel-agent worktree
isolation; boundary-mock fidelity) shipped with v3.1.0.The plugin requires Beads (bd) ≥ 0.47 and jq. The installer fails
fast if either is missing and prints the upgrade command.
# curl-pipe (no clone needed)
curl -fsSL https://raw.githubusercontent.com/preql-data/claude-workflow-plugin/main/install.sh | bash
# or clone-and-run
git clone https://github.com/preql-data/claude-workflow-plugin
cd claude-workflow-plugin
bash install.sh /path/to/your/project
The installer auto-detects v2 layouts (no model: frontmatter, no
.claude-plugin/plugin.json, no MCP servers) and migrates them. You can
also force upgrade mode explicitly:
# auto-detects v2 and migrates
curl -fsSL https://raw.githubusercontent.com/preql-data/claude-workflow-plugin/main/install.sh | bash
# explicit upgrade
curl -fsSL https://raw.githubusercontent.com/preql-data/claude-workflow-plugin/main/install.sh | bash -s -- --upgrade
Migration copies the existing .claude/ to .claude-v2-backup-<timestamp>/
before writing v3 files, so user customizations are recoverable. Run the
diff after install to see what changed:
diff -r .claude-v2-backup-*/ .claude/
irm https://raw.githubusercontent.com/preql-data/claude-workflow-plugin/main/install.ps1 | iex
npx claudepluginhub preql-data/claude-workflow-pluginHarness-native ECC operator layer - 67 agents, 271 skills, 92 legacy command shims, reusable hooks, rules, selective install profiles, and production-ready workflows for Claude Code, Codex, OpenCode, Cursor, and related agent harnesses
Comprehensive feature development workflow with specialized agents for codebase exploration, architecture design, and quality review
Upstash Context7 MCP server for up-to-date documentation lookup. Pull version-specific documentation and code examples directly from source repositories into your LLM context.
A growing collection of Claude-compatible academic workflow bundles. Covers scientific figures, manuscript writing and polishing, reviewer assessment, citation retrieval, data availability, paper reading, literature search, response letters, paper-to-PPTX conversion, and evidence-grounded Chinese invention patent drafting. Rules are organized as reusable skill folders with explicit workflows and quality checks.
Comprehensive skill pack with 66 specialized skills for full-stack developers: 12 language experts (Python, TypeScript, Go, Rust, C++, Swift, Kotlin, C#, PHP, Java, SQL, JavaScript), 10 backend frameworks, 6 frontend/mobile, plus infrastructure, DevOps, security, and testing. Features progressive disclosure architecture for 50% faster loading.
Tools to maintain and improve CLAUDE.md files - audit quality, capture session learnings, and keep project memory current.