plan-orchestrator

Plugin plan-orchestrator from marketplace foundry in piyushsatti/foundry. Commands read from the repo's manifest — 1 month ago.

Skills (8)

• audit

  Independent review of an artifact in scope — structured findings with severity tiers, no fixes. Use when user says audit, independent audit, second opinion, strict auditor, surface parity audit, spec audit of files/plans/PRs (not manifold DB — use manifold skill for MCP spec_audit), regression check

• brief

  Turn a vague or underspecified prompt into a sharp, approved brief before any work starts. Use when the request is unclear, ambiguous, too broad, or the user asks to sharpen, clarify, unpack, or refine the prompt before starting.

• hats

  Use when a plan, design, decision, or document should be reviewed from several distinct professional perspectives before proceeding — the user says different hats, different perspectives, panel review, or names roles to apply; or when a single reviewer's lens would miss whole failure classes (securi

• plan-orchestrator

  Decompose complex multi-component features into a contract-driven DAG of phases with explicit stage gates, transparency protocols, and composition with existing skills. Use when you have a clear what+why+metrics+scope but not a clear how, and the work crosses 2+ components. Refuses to run on undersi

• progress-tracker

  Live dispatch progress for multi-agent orchestration. Use when tracking what dispatched agents are doing, peek_project/peek_phase status, todo_init/todo_done discipline, progress viewer, or "what are agents working on". NOT for spec graph (manifold) or static plans (plan-orchestrator alone). Require

• red-vs-blue

  Use for adversarial stress-testing of a high-stakes artifact — architecture, contracts, launch/no-go decisions, or any plan where a hidden flaw is expensive. Trigger when the user says red team, blue team, attack this plan, or adversarial review. Routine review belongs to audit or hats.

• subset

  Shrink work to the smallest testable vertical slice, define done for that slice, verify end-to-end, then expand. Use when user says take a subset, testable slice, shrink scope, verify end-to-end then expand, or fold it in.

• wardrobe

  Reference library of professional-lens hat files consumed by the consult, hats, and red-vs-blue skills. Not an invocable workflow — consumers read hat bodies by path (../wardrobe/hats/<name>.md). See HATS.md to author or edit a hat.

Subagents (4)

• adjudicator

  Independent adjudicator for the red-vs-blue skill. Dispatch (optionally) as the separate third party that reconciles the red and blue reports into a verdict — never one of the two reviewers. Expects both reports and the artifact in the dispatch prompt. Judges severity from evidence, not from how con

• blue-verifier

  Verifier (blue stance) for the red-vs-blue skill. Dispatch as the other half of the attack/verify pair — it independently confirms what is actually sound with evidence and runs the mandatory coverage check, blind to the attacker. It is NOT a second attacker and never adjudicates red's findings. Expe

• panelist

  Neutral-stance reviewer for the hats skill. Dispatch one per hat — each panelist wears a single wardrobe lens on the same artifact, blind to the other panelists, and returns structured findings. Expects the wardrobe hat file body (frontmatter excluded) and the artifact in the dispatch prompt. Raises

• red-attacker

  Adversarial reviewer (red stance) for the red-vs-blue skill. Dispatch as one half of the attack/verify pair on a high-stakes artifact — it assumes the plan fails in production and hunts for how, blind to the verifier. Expects the artifact (and an optional wardrobe hat file as its lens) in the dispat

MCP servers (1)

• progress-tracker

  runs: python3 ${CLAUDE_PLUGIN_ROOT}/server/mcp_server.py

Similar plugins