Pi Arc Package

Arc issue tracker integration for Pi: packaged Arc skills, prompt templates, session context injection, workflow command aliases, bundled checklist support via @juicesharp/rpiv-todo, and bundled Arc specialist support via pi-subagents.

This package is a Pi-native port of the Claude Code Arc plugin at https://github.com/sentiolabs/arc

What is included

• Prompt templates for common Arc CLI workflows:
  • /arc-create
  • /arc-list
  • /arc-ready
  • /arc-show
  • /arc-update
  • /arc-close
  • /arc-docs
  • and more under prompts/
• Skills for Arc workflows:
  • /skill:arc — general Arc reference
  • /skill:arc-brainstorm
  • /skill:arc-plan
  • /skill:arc-build
  • /skill:arc-debug
  • /skill:arc-review
  • /skill:arc-verify
  • /skill:arc-finish
• Extension commands:
  • /arc-onboard — run arc onboard
  • /arc-which — run arc which
  • /arc-prime — show cached arc prime context
  • /arc-refresh — refresh cached arc prime context
  • /arc-subagents-sync [project|user] — generate Arc specialist definitions from the bundled pi-subagents copy
  • /arc-plan, /arc-build, /arc-review, etc. — friendly aliases for the corresponding skills
• Session context injection:
  • On session start, the extension runs arc prime and injects its output into the system prompt as <arc-context>.
  • Before compaction, the extension refreshes arc prime.
• Bundled @juicesharp/rpiv-todo integration (auto-installed + auto-loaded):
  • todo tool for managing in-session checklist items.
  • /todos command for a quick checklist view/workflow.
  • Persistent overlay widget for visible, session-level task progress.
• Bundled @juicesharp/rpiv-ask-user-question integration (auto-installed + auto-loaded):
  • Provides the ask_user_question tool for Arc workflow decisions using the package questions[] schema.
  • Supports multi-question dialogs, single-select and multi-select questions, optional per-option previews, and per-option notes.
  • Preserves package-provided escape hatches where supported, including Type something. for custom text and Chat about this for returning to free-form conversation.
  • Arc docs should not manually author reserved sentinel labels such as Type something., Chat about this, Other, or Next as normal options.
  • When Arc recommends an option, list it first, append (Recommended) to the label, and explain why in the description.
• arc_agent tool:
  • Runs bundled Arc specialist prompts from agents/*.md in fresh Pi subprocesses.
  • Supports builder, code-reviewer, doc-writer, evaluator, issue-manager, and spec-reviewer.
  • Resolves Arc model tiers (small, standard, large) to concrete Pi models so orchestrators can right-size subagent dispatches.
  • Current limitation: isolation: "worktree" is recognized but not implemented yet.
• Bundled pi-subagents companion support:
  • @sentiolabs/pi-arc bundles and loads pi-subagents by default, so Arc specialists are available without a separate install.
  • Pi's package docs recommend bundling other Pi packages through dependencies + bundledDependencies, then referencing their resources through node_modules/... paths in the package's pi manifest.
  • If you previously installed standalone pi-subagents, remove the standalone package from ~/.pi/agent/settings.json or project .pi/settings.json if duplicate tools or commands appear. The bundled copy from @sentiolabs/pi-arc is enough for Arc workflows.

Prerequisites

• Pi installed.
• The arc CLI available on PATH.
• An Arc project initialized or registered for the working directory:

arc init
# or
arc onboard

The package fails gracefully when arc is unavailable or the current directory is not an Arc project.

Install from npmjs.org

@sentiolabs/pi-arc is published publicly on npmjs.org, so npm installs do not require GitHub Packages registry configuration or a GitHub token.

Install globally through Pi:

pi install npm:@sentiolabs/pi-arc

Install into the current project's .pi/settings.json instead of global settings:

pi install -l npm:@sentiolabs/pi-arc

Pin to a released version:

pi install npm:@sentiolabs/pi-arc@0.1.0

Test without installing permanently:

pi -e npm:@sentiolabs/pi-arc

Install from git

Git installs are supported for source checkouts and unreleased refs:

pi install git:git@github.com:sentiolabs/pi-arc
pi install git:git@github.com:sentiolabs/pi-arc@main
pi install git:git@github.com:sentiolabs/pi-arc@v0.1.0

HTTPS works too if your Git credentials are configured:

pi install https://github.com/sentiolabs/pi-arc

Install locally

From a local checkout:

pi install -l .

Use temporary installation for testing:

pi -e .

Usage

Start Pi in an Arc-enabled project and run:

/arc-onboard
/arc-ready
/arc-create "Fix login bug" --type bug --priority 1
/arc-show <issue-id>
/arc-brainstorm
/arc-plan docs/plans/<file>.md
/arc-build <epic-id>
/arc-finish

You can also invoke skills directly:

/skill:arc
/skill:arc-plan
/skill:arc-build

Arc vs todo boundary

Use Arc for persistent, auditable issue tracking across sessions (arc create, arc update, dependencies, plan shares, and closure history). Use bundled rpiv-todo (todo tool + /todos + overlay) for visible, in-session checklists while you execute the current workflow.

Plan review surfaces

/skill:arc-brainstorm writes design docs under docs/plans/ and asks how to register them for review:

• Legacy local planner: arc plan create <file> for a simple local-only comment thread.
• Encrypted local share: arc share create <file> for encrypted local review with annotations.
• Encrypted remote share: arc share create <file> --remote for reviewers on other machines.

The brainstorm skill writes a first-line marker like <!-- arc-review: kind=share-remote id=<id> -->; /skill:arc-plan reads that marker to choose the matching arc plan or arc share commands.

Arc model tiers

Arc subagents use Pi-native model tiers so the orchestrator can choose a right-sized model per dispatch:

Tier	Default concrete model	Typical use
nano	openai-codex/gpt-5.4-nano	Bulk CLI issue creation and other low-reasoning issue-manager work
small	openai-codex/gpt-5.4-mini	Docs and mechanical edits
standard	openai-codex/gpt-5.3-codex	Normal contained implementation/review
large	openai-codex/gpt-5.5	Complex, cross-cutting, or security-sensitive work

Override the defaults in ~/.pi/agent/settings.json or project .pi/settings.json:

{
  "arc": {
    "modelTiers": {
      "nano": "openai-codex/gpt-5.4-nano",
      "small": "openai-codex/gpt-5.4-mini",
      "standard": "openai-codex/gpt-5.3-codex",
      "large": "openai-codex/gpt-5.5"
    }
  }
}

For compatibility, arc_agent still maps legacy aliases: haiku → small, sonnet → standard, opus → large. New prompts can use nano directly for low-reasoning issue-manager work.

Sync Arc specialists into bundled pi-subagents

@sentiolabs/pi-arc bundles and loads pi-subagents by default, so /arc-subagents-sync refreshes the bundled specialist definitions rather than relying on a standalone install.

Use /arc-subagents-sync to generate Arc specialist agent files from this package's bundled prompts:

• arc-builder
• arc-doc-writer
• arc-spec-reviewer
• arc-code-reviewer
• arc-evaluator
• arc-issue-manager

Pi's package docs recommend bundling other Pi packages through dependencies + bundledDependencies, then referencing their resources through node_modules/... paths in the package's pi manifest.

Bundled resources resolve from ./node_modules/pi-subagents/src/extension/index.ts, ./node_modules/pi-subagents/skills, and ./node_modules/pi-subagents/prompts.

By default, files are written to project scope (<cwd>/.pi/agents/). Pass user or --user to write to ~/.pi/agent/agents/ instead.

If you previously installed standalone pi-subagents, remove the standalone package from ~/.pi/agent/settings.json or project .pi/settings.json if duplicate tools or commands appear. The bundled copy from @sentiolabs/pi-arc is enough for Arc workflows.

The issue-manager agent defaults to the nano model tier and stays phased: create the epic first, then child tasks next, then dependencies/labels after all IDs exist. It prints phase-level timing/progress lines for bulk issue creation. This is sequencing only; true parallel issue creation is not enabled yet.

Generated files include a marker comment so reruns can safely update Arc-managed files while preserving manual edits in user-authored files.

After syncing, verify agent registration:

subagent({ action: "list" })
/agents

Use /subagents-status to monitor active/recent async Arc specialist runs after launch. It does not list idle installed agents.

For Arc gates (especially spec compliance), use Arc specialists (arc-spec-reviewer, etc.) instead of generic worker/reviewer agents.

• Keep arc_agent as the self-contained fallback when Arc pi-subagents definitions are unavailable.
• Claude-style team deployment is intentionally not ported to Pi.

Execution lanes

• Sequential Arc build: use when tasks overlap, dependencies are linear, or pi-subagents is unavailable.
• Parallel Arc batch: use when /arc-plan provides a T0 foundation, file ownership matrix, parallel batch manifest, and validation matrix.
• Ant Colony: future/optional lane for large exploratory work; not a replacement for Arc gates in this iteration.

Naming differences from the Claude plugin

Claude plugin commands used names like /arc:create. Pi prompt templates are filename-based, so this package uses hyphenated names:

Claude plugin	Pi package
/arc:create	/arc-create
/arc:list	/arc-list
/arc:ready	/arc-ready
/arc:show	/arc-show
/arc:plan	/arc-plan or /skill:arc-plan
/arc:build	/arc-build or /skill:arc-build

Current implementation status

Implemented:

• Pi package manifest
• Prompt template migration with /arc-* names
• Skill migration with collision-safe arc-* skill names
• Arc context extension (arc prime cache + system prompt injection)
• Workflow command aliases
• Bundled agent prompt references under agents/
• Bundled @juicesharp/rpiv-ask-user-question package for interactive workflow decisions
• Pi-native arc_agent custom tool for sequential subagent execution
• /arc-subagents-sync command for generating Arc specialist pi-subagents definitions
• Bundled pi-subagents support for worktree-isolated evaluator runs, independent parallel builder batches, and phased issue-manager creation
• Maintainer-only /arc-source-sync workflow for syncing from the Claude Arc plugin source

Not yet implemented:

• Native arc_agent worktree isolation for parallel Arc builders.
• Arc issue autocomplete in the Pi editor.

Intentionally not ported:

• Claude-style team deployment. Pi does not provide Claude's persistent team/task primitives.

Development

Install package dependencies without auto-installing Pi peer dependencies:

npm ci

Run tests and inspect the publish tarball:

npm test
npm run pack:dry-run

Maintainer source sync

This repo includes a maintainer-only /arc-source-sync skill/command for syncing Pi resources from the Claude Arc plugin source while preserving Pi-specific behavior.

Use the skill when asking an agent to evaluate and apply upstream changes:

/arc-source-sync
/arc-source-sync ~/devspace/personal/sentiolabs/agent-nexus/claude-marketplace/plugins/arc

Regenerate migrated resources directly from the default source path:

python3 scripts/migrate-arc-plugin.py

Or pass an explicit Claude Arc plugin source path:

python3 scripts/migrate-arc-plugin.py ~/foo/bar/arc
python3 scripts/migrate-arc-plugin.py --source ~/foo/bar/arc

Smoke test in Pi:

PI_OFFLINE=1 pi -e . --list-models

Useful checks:

rg '/arc:' skills prompts
rg 'TaskCreate|TodoWrite|AskUserQuestion|Claude Code' skills prompts

Release process

Release Please manages package.json, package-lock.json, Git tags, GitHub releases, and CHANGELOG.md. The first official release is bootstrapped to v0.1.0.

Use Conventional Commits on main so Release Please can determine the next version:

feat: add an Arc workflow capability
fix: correct an Arc command edge case

When a Release Please PR is merged, .github/workflows/release-please.yml creates the GitHub release and publishes the package to npmjs.org using npm Trusted Publishing (OIDC) with provenance.

Before the first automated npm release, configure npm Trusted Publishing for the package:

• Package: @sentiolabs/pi-arc
• Publisher: GitHub Actions
• Repository owner/user: SentioLabs
• Repository: pi-arc
• Workflow filename: release-please.yml
• Environment: leave blank unless a GitHub Environment is intentionally required

npm Trusted Publishing requires an existing package to configure from package settings. If @sentiolabs/pi-arc has not been published before, publish a one-time public 0.0.0 bootstrap version or configure the trusted publisher with npm trust after the package exists, then let Release Please publish the first official v0.1.0 release.