claude

polaris-team-os

Polaris: 27 skills for the delivery cycle — base-workflow routes each request to grill, prototype, plan-area, implement-plan, tdd-dev, bugfix, review, real-testing, ship and the rest; start and end keep the session. Start with docs/TEAM-ONBOARDING.md.

0 on GitHubby Niccolò Franzinv0.35.0MITupdated 19 days ago

Not in a marketplace yet, so it loads for one session with --plugin-dir.

Plugin polaris-team-os in legacynik/polaris. Commands read from the repo's manifest — 14 days ago.

This plugin runs code on your machine — it starts local MCP servers. The exact commands are listed below; read them before installing.

What it installs

Skills (27)

  • architecture

    ◇ Polaris — Decide one architectural commitment and record it, usually when the move is already picked and offered for a sanity check. Fires when a change moves a boundary, an interface, a data contract or a dependency, when a schema changes under separately deployed readers, when a query is fast un

  • babysit

    ◇ Polaris — Use after a PR is opened or pushed, and when asked to babysit, watch, check on or get a PR green, until it carries the owner's label on its current head. Never merges or approves. Requires gh and python 3.

  • base-workflow

    ◇ Polaris — The router. On any request that touches code or the tracker (an idea, "add X", "fix Y", a traceback, "open the PR", "track this"), name the case, invoke the skill that owns it, and at the end of each skill propose the next one or stop and ask with the options. Use at the start of work an

  • bugfix

    ◇ Polaris — Close a bug through 12 stages, each passed by an exit code — find without assuming, fix with TDD, prove the consequence, rewrite the tests that lied — with a fix-loop breaker and side findings filed as beads. Use when a traceback, a bug issue, a red smoke, a log error or "this doesn't wo

  • crasher

    ◇ Polaris — Drive a deployed app like a user, hunt failures, and PROPOSE one deduplicated finding per confirmed crash for a person to pick, never filing on its own; or judge whether a run's no-crash verdict stands. Use for does staging still work, smoke the deployment, hunt regressions on the live a

  • critic

    ◇ Polaris — Attack one artifact from fresh context: a plan before it is accepted, a decision record, a claim, or a diff that needs more than review. Only quoted, falsified findings survive; blocking ones go back to the author, only taste reaches the human. Use before accepting a plan or decision, or

  • database-schema

    ◇ Polaris — Read the real schema before any database work — locate migrations, ORM models and generated types, verify every table and column you will touch, and refuse a stale schema reference as truth. Use before writing or changing a query, a migration, a model, an RLS policy, a seed or a data fix

  • domain

    ◇ Polaris — Name the domain and place its boundaries. Fires when a term is overloaded, when two words are used for one thing, when a subdomain must be typed core, supporting or generic, or when two contexts with different models must integrate or be mapped. Not for where a service boundary falls, no

  • end

    ◇ Polaris — save the session, as a checkpoint or as the close. Every call appends a session-log entry, rewrites your lane line in the cockpit, records lessons, lists decisions with their homes and commits pathspec-only; the close also prunes the cockpit against git and the tracker and leaves a hando

  • eval

    ◇ Polaris — Measure prompts, agents and skills before calling them better. Build an eval suite from real failures, choose between a golden set, code checks and a judge, calibrate the judge, audit an existing eval, run a bounded keep-or-revert loop, prove a skill change test-first, gate a release or

  • git-power

    ◇ Polaris — Use before any Git commit or PR boundary and before any destructive Git action. MUST invoke for every `reset --hard` and every pre-PR comparison of a branch against origin/main; never substitute raw Git. Also censuses and safely closes worktrees, branches, stashes and PRs. Triggers on co

  • grill

    ◇ Polaris — Discover, phase 1. Grill an idea against the repo's docs, earlier ideas, how other repos solved it and what is in or out of scope; list every assumption with how to test it; propose the size; open the plan file at status discovery. Use for "grill this idea", "clarify this feature", "chal

  • implement-plan

    ◇ Polaris — Build an accepted plan's Beads epic phase by phase: claim the next ready task, hand it to a fresh-context agent that records RED first, close each phase only through the feature gate. Use when a plan is accepted and its epic poured, or to resume a build after a crash. Requires bd (Beads)

  • init

    ◇ Polaris — bring an existing repository into the harness: audit, propose, then write a typed CONTEXT/ARCHITECTURE/DOMAIN backbone; with --team required|optional, pin a Polaris release the whole team auto-installs. Use when a repo has none of those or the context pack emits nothing there. Not for re

  • issue-contract

    ◇ Polaris — Create or update GitHub issues with a clear definition of done and regression checks, and route operational work to a Beads bead instead. Use when asked to open an issue, track a bug or finding, promote a bead, or hand work to another lane. Requires bd (Beads), gh and python 3.

  • plan-area

    ◇ Polaris — Write ONE plan file per piece of medium or large work (area first, outcome, DoD, contracts, every claim assumed or proven), pour its phases into a Beads epic gated by feature_gate.py, record course changes, and close it after merge. Use when an idea is scoped and needs a plan, the plan m

  • pr

    ◇ Polaris — Use when writing, retitling or rewriting a pull request body, before `gh pr create` or `gh pr ready`, so its owner judges it merge-ready on GitHub without reading code: the verdict on top, every issue criterion answered with proof, the smallest visual, before/after evidence and merge ris

  • problem-solver

    ◇ Polaris — Unstick a problem by changing how it is looked at: name what repeats and the crux, propose three opposite ways of thinking about it, let the founder pick, then work the problem through that lens to an answer and the cheapest experiment that confirms it. Use after repeated failed attempts

  • prompt-engineering

    ◇ Polaris — Write or change the words a model reads — system prompts, subagent dispatches, judge text, hook and deny messages, questions for typed decision models. Use when a prompt must be written, fixed or tightened; hand the candidate to `eval` to measure it, never done without that measure. Not

  • prototype

    ◇ Polaris — Prototype, phase 2. Answer one named design question with throwaway code on a spike branch — does this state model or logic hold, what should this UI look like — and write only the verdict into the plan file. Use when a plan assumption is still assumed and code can settle it, or before b

  • real-testing

    ◇ Polaris — Verdict from the real product: serve the commit, walk the customer journey through the UI on every surface it reaches, let a fresh agent try to break it, leave the tests in CI. Modes: local, merge (receipt bound to the tree), production check. Use before merge or promotion, on 'test the

  • review

    ◇ Polaris — The audit of a change on six axes: tests prove behaviour, no drift from governing decisions, constraints kept, the plan's DoD met (no less, no more), simplicity and reuse, security. Use on the whole diff before ship, or before a commit or merge; second-opinion mode asks another model fam

  • ship

    ◇ Polaris — Use when the work is done: one command that tests, audits, real-tests, updates the docs the diff made stale, opens the PR, signs it on staging once, writes the body and drives it to the owner's label, then merges once on the owner's yes and closes the work. Triggers: ship it, open the PR

  • start

    ◇ Polaris — resume the current objective from the latest local handoff, read-only and local, and open the session that end checkpoints and closes; first use also provisions your own team path. Use at session start or on "start / resume / where were we". Not a status view. Requires git, python3; gh o

  • status

    ◇ Polaris — the explicit, full situational view of a portfolio or one repository — live git and GitHub pulse, active outcomes, team collisions, freshness, decisions accepted and still not done, and the one highest-leverage next move. Use on "polaris status / status completo", cross-repo priorities,

  • tdd-dev

    ◇ Polaris — Drive one behaviour change test-first — a hash-bound RED, the minimum fix, a break proving the test guards the code, then the repo's regression suite. Use when implementing a feature or behaviour change, fixing a reproducible bug through a regression test, proving an existing fix's test,

  • week

    ◇ Polaris — your own week, both ends. At the start, build your weekly focus from carryover and live tracker issues; after a finished Sunday→Saturday week, report it against that focus with merged-PR evidence. Both open with the decisions accepted and still not done. Use on "plan my week / what am I

MCP servers (2)

  • sequential-thinking

    runs: npx -y @modelcontextprotocol/server-sequential-thinking

  • serena-code-graph

    runs: uvx --from git+https://github.com/oraios/serena@16e8e24da103d47beb5f43df1f5585985f9aabd6 --with pyright python ${CLAUDE_PLUGIN_ROOT}/scripts/serena_code_graph_launcher.py

View source on GitHub

Problems found in this repo's manifests

  • .claude-plugin/marketplace.json is not valid JSON

Similar plugins

superpowers

claudecodex

Core skills library for Claude Code: TDD, debugging, collaboration patterns, and proven techniques

15 skills · hooks

obra/superpowers 316,836

claude-mem

claudecodex

Persistent memory for Claude Code: the next session starts where the last one left off.

29 skills · 1 MCP server · hooks

thedotmack/claude-mem 116,618

mem0

claudecodex

Cross-session memory and token savings for coding agents.

6 skills · 1 agent · 1 MCP server · hooks

mem0ai/mem0 68,893

Backend API design, GraphQL architecture, workflow orchestration with Temporal, and test-driven backend development

9 skills · 1 command · 8 agents

wshobson/agents 42,461

agentmemory

claudecodex

Persistent memory for AI coding agents -- captures tool usage, compresses via LLM, injects context into future sessions

17 skills · 1 MCP server · hooks

rohitg00/agentmemory 30,326

beads

claudecodex

AI-supervised issue tracker for coding workflows

1 skill · 1 agent · hooks

gastownhall/beads 28,698

Maintain polaris-team-os?

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 pluginslisted