agent-workforce

Plugin agent-workforce from marketplace agent-workforce in jayheavner/agent-workforce. Commands read from the repo's manifest — 1 month ago.

Skills (24)

• agent-workforce

  Route a task through the AI Agent Workforce's architect, builder, debugger, verifier, reviewer, deployer, researcher, operations, scribe, and ticketing roles with proportional unattended execution, narrow decision gates, repair loops, and evidence-based closeout. Use only when the user explicitly in

• auditing-requirements

  Audit a business-requirements document for standards violations — implementation detail, technology specifics, architecture rationale, or test procedures that belong elsewhere — and report each with severity and a recommended destination. Use after authoring requirements, before requirements code re

• close-ticket

  Close an Asana subtask against its acceptance criteria with evidence, using the verifying discipline — all criteria pass or the task stays open. Use before marking any Asana task or subtask complete.

• closeout

  Finish a workforce task the same way every time — fresh verification, a focused commit, one status note, honest completion claims, and the exact whole-session cost report. Use when work is being finished, shipped, wrapped up, handed off, or reported complete.

• convene-panel

  Convene a panel of named experts against an artifact — discussion, debate, or socratic mode — with required per-expert dissent and a five-section synthesis that never averages disagreement. Use when the user asks for a panel, an expert review of a spec/plan/document, or a structured multi-perspectiv

• debugging

  Diagnosis loop for real bugs — build a red-capable feedback loop first, minimise the repro, rank falsifiable hypotheses, instrument one variable at a time, fix with a regression test, clean up. Use when encountering any bug, test failure, or unexpected behavior, before proposing fixes.

• finishing-a-branch

  Take completed branch work from "code done" to integrated and cleaned up — fresh verification gate, atomic commit of only this work, integration path chosen by the human, post-merge cleanup. Use when implementation is complete and the work needs committing, merging, or a PR; when asked to "wrap up",

• growing-the-team

  Create a new skill or agent for the workforce when a task exposes a capability gap — draft it in the workforce repo marked provisional, use it immediately for the live task, and disclose it at closeout for human review and possible upstreaming. Use when the practitioner test fires (a practitioner of

• handing-off

  Write a handoff brief that lets a fresh session resume mid-task work without re-derivation — exact next actions, verification state split into proven vs unrun, dirty-tree disclosure, decisions with their why. Use when ending a session mid-task, when asked to "hand off", "wrap up for the day", or "wr

• handling-secrets

  Universal secret-handling discipline — never materialize a credential in any file, log, or transcript; reference via env vars or on-demand reads. Use when any task touches credentials, keys, tokens, or .env files.

• interviewing

  Interview the human about a design one decision at a time until shared understanding is confirmed — facts from the codebase, decisions from the human. Use when a request is fuzzy, before writing any spec or plan, or when the user asks to be grilled about a design.

• onboard-project

  Declare a project's issue tracker and tool ready-checks in .workforce/project.json so every session starts on checked facts instead of guesses. Use when the session-start probe or cost report says the tracker is UNDECLARED, when a project needs its required tooling captured, or when the user asks to

• op-migration

  Migrate credentials discovered by the handling-secrets discipline into 1Password using service-account authentication only. Use when moving exposed secrets into a 1Password vault, or when a task needs an op:// reference for a credential.

• planning

  Write implementation plans an engineer with zero context can execute — exact files, complete code in every step, bite-sized TDD tasks, checkbox tracking, self-review. Use when turning an approved spec into tasks, before touching code.

• project-policy

  Jay's org policy values — thresholds, formats, and tool mandates consulted by framework skills. Use when any framework skill resolves a policy:<key>.

• refine-request

  Assess an incoming request before work starts — a four-check gap scan (deliverable, scope, done-condition, authority) that asks only about what is materially missing, so complete requests proceed untouched and lazy prompts get one sharp round of questions. Use at intake of any new request, when a pr

• review-ticket

  Review an Asana ticket for ownership, validity, and true scope before starting work on it — the mandatory reconnaissance step before any tracked work begins. Use when given a ticket ID or URL, asked "what should I work on?", or before starting any Asana-tracked work.

• reviewing

  Code-review craft in both chairs — review a change against project standards and its originating requirement (spec fidelity) with severity triage and locations + fixes, and receive review findings on your own work with verification before implementation. Use when reviewing code, a pull request, or a

• tdd

  The red→green loop and what makes it produce tests worth keeping — seams, anti-patterns (tautological, implementation-coupled, horizontal slicing), rules of the loop. Use when implementing any feature or bugfix, before writing implementation code.

• ux-to-ui-design

  Use when designing or modifying any UI — panels, layouts, navigation, forms, editors, dashboards, components, interaction patterns, or visual hierarchies. Use before proposing, planning, or implementing UI changes, including when reviewing UI work. Apply when a task involves screens, modes, states,

• verifying

  Evidence before claims — check each stated acceptance criterion by running the exact command and reading its real output; per-criterion verdicts pass/fail/UNCHECKED with evidence attached, never averaged. Use when about to claim work is complete, fixed, or passing against enumerable criteria — closi

• write-ticket

  Create well-structured Asana tickets and subtasks using separation of concerns, objective acceptance criteria, and correct dependency sequencing. Use when drafting a new ticket, decomposing work into subtasks, or setting task dependencies for tracked work.

• writing-business-requirements

  Use when writing or reviewing business requirements documents following industry standards (BABOK v3, IEEE 830). Use when drafting new functional requirements, fixing requirements defects found in an audit, or extracting business needs from mixed documentation.

• writing-skills

  Author or edit a skill for this framework — eval scaffolding written first, a description that routes correctly, a policy-free body per the house standard, admission only through a run record. Use when creating a new skill or pack, editing an existing skill's content or description, or reviewing a s

Subagents (13)

• architect

  Designs systems, writes specs and implementation plans for the agent team. Also drafts new skills and agents when the team grows itself. Dispatched by the orchestrator; not for direct casual use.

• builder

  Implements code using TDD — from a reviewed plan or, for contained work, directly from a well-scoped dispatch. Dispatched by the orchestrator; not for direct casual use.

• debugger

  Diagnoses symptoms — broken behavior, failing systems, "why is X wrong" — and returns a root cause with evidence. Dispatched by the orchestrator before any fix is routed; not for direct casual use.

• deployer

  Executes authorized cloud deployments (SAM, Amplify, CDK) with smoke checks and rollback discipline. Dispatched by the orchestrator; not for direct casual use.

• executor

  General-purpose shell runner for authorized work — arbitrary commands, installs, and shell-level file operations. Never authors source or tests; that is builder work. Dispatched by the orchestrator with the stated intent; not for direct casual use.

• ops

  Investigates and administers AWS, Azure, and Okta. Cloud reads run freely; authorized mutations execute within the dispatched scope. Dispatched by the orchestrator; not for direct casual use.

• orchestrator

  Team lead for multi-phase orchestrated work. Use ONLY when the user explicitly asks for the orchestrator or the agent team. Intended to run as the main session, started via the agent-workforce launcher (never plain `claude --agent orchestrator`, which drops permissions and freshness), not as a dispa

• researcher

  Investigates questions across the web, Glean, and codebases; returns cited findings. Dispatched by the orchestrator; not for direct casual use.

• reviewer

  Reviews code for quality and security, and critiques plans and specs when dispatched in those modes. Dispatched by the orchestrator; the dispatch names the mode. Not for direct casual use.

• scribe

  Writes documents — reports, design briefs, business requirements, postmortems, and the team's per-task status notes. Dispatched by the orchestrator; not for direct casual use.

• test-author

  Writes the acceptance test suite from a reviewed plan BEFORE the builder implements — a separate author for the tests the code must pass. Dispatched by the orchestrator on design routes; not for direct casual use.

• ticketer

  Writes, reviews, and tracks Asana tickets per the org's ticket disciplines. Dispatched by the orchestrator; not for direct casual use.

• verifier

  Runs test suites and validates acceptance criteria with evidence. Dispatched by the orchestrator; not for direct casual use.

Hooks (9)

• PreToolUse

  "${CLAUDE_PLUGIN_ROOT}/hooks/agent-team-plugin-router.sh" secrets

• PreToolUse

  "${CLAUDE_PLUGIN_ROOT}/hooks/agent-team-plugin-router.sh" dispatch

• PostToolUse

  "${CLAUDE_PLUGIN_ROOT}/hooks/agent-team-plugin-router.sh" audit

• PostToolUse

  "${CLAUDE_PLUGIN_ROOT}/hooks/agent-team-plugin-router.sh" cost

• SubagentStop

  "${CLAUDE_PLUGIN_ROOT}/hooks/agent-team-plugin-router.sh" subagent-stop

• Stop

  "${CLAUDE_PLUGIN_ROOT}/hooks/agent-team-plugin-router.sh" closeout-stop

• Stop

  "${CLAUDE_PLUGIN_ROOT}/hooks/agent-team-plugin-router.sh" archive-run

• SessionEnd

  "${CLAUDE_PLUGIN_ROOT}/hooks/agent-team-plugin-router.sh" archive-run

• SessionStart

  "${CLAUDE_PLUGIN_ROOT}/hooks/agent-team-plugin-router.sh" session-start

Similar plugins