helmsman-ai

helmsman-ai/helmsman-claude-plugin

Plugin Claude Code

Agentic SDLC workspace for Claude Code — guides developers from PRD to shipped code with a durable artifact dossier.

helmsman-ai/helmsman-claude-plugin

Plugin Claude Code

Agentic SDLC workspace for Claude Code — guides developers from PRD to shipped code with a durable artifact dossier.

helmsman-ai/helmsman-claude-plugin

Instructions file

Claude Code instructions for helmsman-ai/helmsman-claude-plugin, covering claude.md, what this repo is, how the plugin fits together, skill anatomy and gates.

helmsman-ai/helmsman-claude-plugin

Agent

Pragmatic architect who produces tech designs with evaluated alternatives and recorded ADRs (Stage 04), then breaks the approved design into atomic, dependency-ordered implementation tasks (Stage 05). Invoked by the Orchestrator for both stages.

helmsman-ai/helmsman-claude-plugin

Agent

Disciplined coder who implements one task at a time, writes tests first, follows repo conventions, commits with structured messages, and updates the progress log. Invoked by the Orchestrator at Stage 06, once per task.

helmsman-ai/helmsman-claude-plugin

Agent

Helmsman's central coordinator. Routes work to specialist sub-agents, enforces quality gates, manages project state, and maintains the decision log. Invoke for any Helmsman slash command or when the user interacts with an active project.

helmsman-ai/helmsman-claude-plugin

Agent

Skeptical senior engineer who critiques raw PRDs and produces clean, implementable specifications. Invoked by the Orchestrator at Stage 02 (PRD Review & Gap Filling). Never invoked directly by the user.

helmsman-ai/helmsman-claude-plugin

Agent

Investigative agent that maps the existing codebase, identifies prior art, builds a stakeholder map, and surfaces dependencies before the architect designs anything. Invoked by the Orchestrator at Stage 03 (Discovery & Research).

helmsman-ai/helmsman-claude-plugin

Agent

Picky senior reviewer who checks each implemented task against its acceptance criteria, conventions, security requirements, and edge cases. Produces a structured review report with severity-rated issues. Invoked by the Orchestrator at Stage 07, once per implemented task.

helmsman-ai/helmsman-claude-plugin

Command

Run the gate checklist for the current stage, then invoke the next stage's specialist agent. Requires the current stage to be in-review or complete. Blocks on hard gate failures.

helmsman-ai/helmsman-claude-plugin

Command

Mark the current stage (or a specific artifact) as accepted by the developer. Updates state.yaml and appends to the decision log. Does not automatically advance — run /advance after /approve to trigger the next stage.

helmsman-ai/helmsman-claude-plugin

Command

Add inline feedback on the current stage's output and re-invoke the responsible agent to address it. The agent revises artifacts in response to the comment. Does not change stage status.

helmsman-ai/helmsman-claude-plugin

Command

After a project is complete (or at any stage), analyze project artifacts and propose memory updates across all three tiers: global, repo, and project. Each proposal is shown to the user for explicit approval before any file is written.

helmsman-ai/helmsman-claude-plugin

Command

Compile a fresh project dossier from all approved artifacts. Produces projects/ /dossier.md — the single readable record of the project for future maintainers. Can be run at any stage; sections not yet completed are marked as in-progress.

helmsman-ai/helmsman-claude-plugin

Command

First-run setup wizard. Creates the workspace directory structure, writes manifest.yaml, scaffolds memory/ and templates/ directories, registers the first repository, and optionally initializes git and recommends the GitHub connector. Safe to re-run — will not overwrite existing data.

helmsman-ai/helmsman-claude-plugin

Command

Move the project's currentstage backward to a specified stage. Marks all downstream stages as stale. Does not delete artifacts. Logs to decision log.

helmsman-ai/helmsman-claude-plugin

Command

Change a gate's severity for this project, or bypass a failing gate once. Writes to state.yaml and the gate audit log. Requires a reason.

helmsman-ai/helmsman-claude-plugin

Command

Quick-reference dashboard of all Helmsman projects: name, mode, current stage, status, and direct-action commands. Lighter than /status — one line per project, no artifact details. Read-only.

helmsman-ai/helmsman-claude-plugin

Command

Re-run all stale stages in pipeline order. For each stale doc stage: snapshot, re-run specialist agent, diff, present for approval. For the implementation stage: produce a Change Impact Report instead. Never auto-reruns implementation.

helmsman-ai/helmsman-claude-plugin

Command

Display a timeline of all jump-backs, propagations, snapshot operations, and CIR events for the current project. Read-only — changes nothing.

helmsman-ai/helmsman-claude-plugin

Command

List snapshots for the current project, or restore a specific snapshot. Snapshots are taken automatically before each propagation re-run.

helmsman-ai/helmsman-claude-plugin

Command

Create a new Helmsman project: scaffold the directory structure, save the PRD input, initialize state.yaml, and immediately kick off the PRD Reviewer agent.

helmsman-ai/helmsman-claude-plugin

Command

Show the current state of one or all Helmsman projects: active stage, gate status, pending approvals, recent history, and gate override summary. Read-only — changes nothing.

helmsman-ai/helmsman-claude-plugin

Skill Claude CodeCodex

Security-hardened code review. Extends the standard review with mandatory threat model verification, dependency audit, and secrets scanning. Intended for teams shipping customer-facing or security-sensitive features.