claude

jplugin

Spec -> plan -> TDD build -> review -> wrap-up workflow skills for coding agents; the plugin is this repository.

1 on GitHubby jplugin-agentic-development maintainersv1.8.8MITupdated 3 days ago
/

Plugin jplugin from marketplace jplugin-agentic-development in Joaovsales/jplugin-agentic-development. Commands read from the repo's manifest — 2 days ago.

This plugin runs code on your machine — its hooks fire automatically on events like tool use. The exact commands are listed below; read them before installing.

What it installs

Skills (44)

  • auto-push

    Semi-autonomous pipeline. User describes an idea; agent runs /plan and PAUSES for explicit approval. After approval, /build and /wrap-up-session run autonomously through commit and push.

  • brainstorm

    Explore a feature idea through divergent design thinking before committing to a spec. Use before /plan for non-trivial features requiring design decisions.

  • checkpoint

    Snapshot current session progress to tasks/checkpoint.md for handoff or pause.

  • create-verification-skill

    Generate a project-local verification skill that drives the real app through its user surface. Use when a repository has no grounded way to prove UI, CLI, desktop, API, mobile, or library behavior.

  • debug

    Systematically investigate, diagnose, and fix bugs using root cause analysis. Use when debugging errors, test failures, runtime issues, or when the user reports a bug. Integrates with the typed learning store (tasks/solutions/).

  • design-stack

    Set up, update, and browse the shared UI design sources used by jplugin. Installs Taste, Impeccable, design references, img2threejs and HyperFrames once in user scope.

  • diagram

    Draw a checked, explorable standalone HTML diagram of a system, flow, schema or history through the installed Archify tool — architecture, workflow, sequence, dataflow, lifecycle, ERD, class, tree, timeline or waterfall. Use whenever the user wants a picture of how something is built or runs: 'draw

  • eval

    Blinded A/B evaluation of a skill, prompt, or workflow change before promoting it. Candidates run in sanitized worktrees on organic prompts and never learn they are being measured; grading comes from session transcripts, not self-report. Three runnable modes: triggerability (does the harness route t

  • folder-context-optimization

    Sweep a folder to identify legacy/unused files, propose archival, and update docs. Use when a directory feels bloated or disorganized.

  • go

    Natural-language front door for interactive work. Give it a goal in plain words — a question about the code, a defect, a rename, a slowness, an open PR to drive green, or a feature — and it picks one lane from the shared lane catalogue, prints a [ROUTE] line, records the lane's steps in tasks/todo.m

  • grill-me

    A relentless interview to sharpen a plan, decision, or idea, typed by the user and never started by the agent. Use when the user says 'grill me' about something.

  • grilling

    Interview the user relentlessly about a plan, decision, or idea until a shared understanding is reached, working the design tree in frontier rounds with a recommended answer on every question. Use when the user wants to stress-test their thinking, uses any 'grill' phrasing, or when another skill nee

  • how

    Use for \"how does X work\", code walkthroughs before changing something, and placement / ownership / layering questions (\"where should this live\", \"which package owns this\", \"is this the right layer\"). Explains subsystem architecture, runtime flow, onboarding mental models. Use why for motiva

  • html-presentation

    Generate a polished, self-contained HTML presentation (report or slide-deck) from structured content. Use when another skill or the user needs to publish a session review, design audit, project summary, or any narrative as a beautiful HTML document with strong visual and information-design quality.

  • learn

    Extract durable learnings from the current session and persist them as typed documents in tasks/solutions/.

  • maintain-verification-skill

    Reconcile a project verification skill after changed user behavior or run a full source-and-live feature audit. Use after user-facing changes or when auditing a verify-app feature map.

  • make-it-simpler

    One behavior-preserving simplification a day, sized to one review and shipped as a pull request. Ranks the tree with scripts/signals.py (a rule stated in many files, citations that no longer resolve, always-loaded prose over budget, docs contradicting their scripts, prose a test already enforces, or

  • memory-maintain

    Sweep the typed learning store (tasks/solutions/) — resolve needs_review documents, merge duplicates, prune stale or contradicted entries. Invoked at every session start and wrap-up; self-gates on session count.

  • plan

    Interview user, write a feature spec, slice it into session-sized build steps, and end with a build prompt for a fresh session. Use for any non-trivial feature before coding.

  • prd

    Interview the user about a greenfield project, produce a structured PRD, ordered backlog, and agent context file. Use as the entry point for new projects.

  • quality-gate

    The post-build review gate: run when all tasks in tasks/todo.md are done, before wrap-up or commit. Sequential phases — structural quality and reuse (simplify), AI anti-pattern cleanup (deslop), the security checklist, and APOSD design audit — emitting four-axis findings (severity, confidence, autof

  • receive-review

    Process incoming code review feedback with technical rigor. Use when receiving review comments on PRs, from users, or from automated review tools.

  • refresh

    Context reset for a session running out of room: snapshot working state to disk, then continue the work in a fresh context rebuilt from that snapshot. Distinct from /checkpoint, which saves progress and stops — reach for /refresh when the work continues but the context must be recycled. Triggers on:

  • security-scan

    OWASP-focused security audit scoped to the files changed in this session, worked through an explicit per-category checklist (input validation, authn/authz, secrets, crypto, dependencies, error handling). Use before committing or deploying changed code. Triggers on: 'security scan', 'check for vulner

  • setup-deployment

    One-time interactive bootstrap for deployment verification. Scans the project for deployment signal files, asks the user to confirm detected services and project IDs, and writes the routing table into AGENTS.md below the managed block.

  • slice

    Break a spec's acceptance criteria and implementation paths into session-sized slices — a Build Order table, a plan block in tasks/todo.md, the rendered visual plan page and a build prompt — and the one dry-run that mints the feature row's id. Use after a spec carries § Decisions and § Acceptance Cr

  • software-design-expert-learn

    End-of-session code review and design tutorial based on 'A Philosophy of Software Design' by John Ousterhout. Use when the user wants to review all code written in the current session, understand design trade-offs, learn software design principles, or improve code review skills. Triggers on: 'review

  • software-design-expert-review

    Run a focused APOSD design review on recently changed files. Scans for the 10 red flags from 'A Philosophy of Software Design' plus Error Design (R11), emits four-axis findings (severity, confidence, autofix_class, owner), and produces a GO / HOLD / STOP verdict. Can be invoked manually or called by

  • start-qa

    Discover project config, restart app, and launch browser for manual QA testing.

  • sweep

    Producer routine engine — verify the codebase through one lens (janitor for bugs, architect for design) and file every proven finding as a registry task with a reproduction and a proposed fix. Use unattended from the janitor and architect routines, or by hand for a weekly audit.

  • sync

    Pull latest skills, hooks, agents, and config from the jplugin-agentic-development template repo.

  • system-design-planning

    Turn an issue or a feature idea into an upstream architecture review — system design, component contracts, data models, constraints, and a dependency-ordered build plan — rendered as a self-contained HTML document, then ends with a build prompt for a fresh session. Nothing is filed and nothing is bu

  • task-registry

    Resolve one task against an external tracker (GitHub Issues) or a local Markdown store. Use when reading a task's full record, recording discovered work as a task, checking which tracker is configured, or selecting and claiming the next issue for a routine.

  • tdd

    Red-green-refactor for a list of acceptance criteria or a bug's expected behaviour, with Prove-It for bugs (reproduce with a failing test first). Called by /build for each slice and by /debug for its fix step; use directly when asked to do TDD or to write the test first.

  • tidy

    Harness hygiene sweep for a jplugin-agentic-development repository — the template, its mirror, and any project that vendored the harness through /sync. Nine checks over the surfaces that duplicate by design (skills tables, retired skills, installed copies under ~/.claude and ~/.agents, backticked pa

  • verify-deployment

    Wait for post-push deployment builds to resolve, fetch logs on failure, and loop a code-debugger fix cycle up to 3 iterations before escalating. Service-agnostic — driven by runbook files in .claude/deployments/.

  • verify-evidence

    Enforce evidence-based verification before any completion claims. Supports --scope deployment and --scope e2e. Use before committing, creating PRs, marking tasks done, or claiming success.

  • verify-task-registry

    Verify this repository's task-registry CLI through isolated local task creation, reading, and routine selection with retained PTY evidence.

  • visual-plan

    Render an existing text spec into a self-contained HTML visual plan — typed cards for decisions, criteria, risks, open questions and slices, real diagrams, a blockers panel, the build prompt and a review export — for approval before implementation. Use after /plan has already written specs/<feature>

  • visual-recap

    Turn a completed branch's git diff into a self-contained HTML visual recap — narrative, file-tree, and annotated key changes. Use after implementation to produce a richer review artifact than a plain diff.

  • why

    Use for 'why does X work this way', 'why we picked Y', design rationale, regressions, postmortems, or data-backed thresholds. Discovers available MCPs and queries each evidence category (source control, issue tracker, long-form docs, real-time chat, infrastructure observability, error tracking, prod

  • wrap-up-session

    Close the session — learnings, spec reconciliation, the quality-receipt check, the full suite, then a clean commit, push and pull request. Use at the end of any coding session.

  • writing-skills

    Author new skills with proper structure, iron laws, and reference docs. Use when creating or improving skills for the workflow.

  • yolo

    Fully autonomous loop. User describes an idea; the agent runs /plan, /build, and /wrap-up-session in a Ralph-style loop until the backlog is empty or a circuit breaker trips. No user prompts between phases.

Hooks (3)

  • SessionStart

    bash "${CLAUDE_PLUGIN_ROOT}/.agents/hooks/session-start.sh"

  • PreCompact

    bash "${CLAUDE_PLUGIN_ROOT}/.agents/hooks/pre-compact.sh"

  • Stop

    bash "${CLAUDE_PLUGIN_ROOT}/.agents/hooks/session-stop.sh"

View source on GitHub

More from Joaovsales/jplugin-agentic-development

All 3 plugins →

Kanban pane of the sub-agents in this session: Plan / Build / Review / Human / Done, one clickable card per agent.

1

Claude Code mod: one docked Session pane showing the workflow phase, slice progress, the sub-agent tree with drill-down and the quality gates of a jplugin session. Observe-only; install id session-viz@jplugin-agentic-development.

1

Maintain jplugin?

This page updates itself from your repo. Add the badge to your README, or get it Featured to pin it above the organic results with a followed link to your site.

claude & codex plugins3 plugins