build
claudeBuild the work: a test-first loop, worktree setup, and fan-out of a spec's issues to herdr tabs or T3 Code threads.
31 skills
Ship issue work: commit and close, commit and open a PR, work through reviewer feedback, and write release notes.
Plugin ship from marketplace agent-skills-kit in devarfeen/agent-skills-kit. Commands read from the repo's manifest — 8 days ago.
agentic-qaAgent-run functional QA of a PR or branch that reaches a screen, incl. a PR /factory puts in QA: runs acceptance criteria in the live app, records the result. Never edits code. Pixels: /pixel-audit.
agents-mdGenerate workspace-root AGENTS.md and rules from .code-workspace, or sync project runtime blocks. Never writes CLAUDE.md/GEMINI.md; a request to make a shim, even with AGENTS.md: writing-for-agents.
ask-kitRoute kit skill-map questions; recommend one skill, never run it. Excludes install searches and worker dispatch. Existing spec, issue, ticket, or PR: /factory. Matt-only routing: ask-matt.
ci-loopDrive an open PR's failing CI to green: read the log, reproduce, fix in scope, push, up to 3 attempts under one approval. Reviewer comments are /pr-feedback; no PR or CI run is /diagnosing-bugs.
commit-push-closeShip finished work, even without an issue: structured commit, push, then close the issue with how-to-test steps. Use for commit, push, and close, or bare ship it. A reviewable PR is /commit-push-pr.
commit-push-prShip issue work as a pull request: structured commit, push, PR with Closes #N. Use only when the user explicitly asks for a PR; a bare ship it is /commit-push-close.
deploy-watchWatch a merged PR's staging deploy, smoke-check staging when approved, and record pass or fail on the PR; answers \"did it reach staging?\". Never touches a server or deploys. Broken: /staging-fix.
design-systemOnce per UI project: turn a design system (Figma, spec, reference screens, or a guided definition when no designs exist) into tokens, a UI library, a preview page, and an AGENTS.md rule; extend later.
factoryShows where a spec (PRD), ticket, or PR stands in the factory workflow and names the one skill that moves it on. Read-only unless the user starts /factory run. Never closes issues or passes staging.
feature-discoveryTrace how a named feature, module, workflow, API, config, or symbol works and who uses it; read-only. Whole-repo architecture summaries: /teach. Porting requires a separate request.
feature-promptTurn a feature idea, change request, or rough requirement into a small prompt for grill-with-field. An ADR, even as a prompt, is /to-spec; how existing behaviour works is /feature-discovery.
grill-with-fieldGrill a plan with field research: fetched standards, provider rules and proven tools become grilling questions; `check` audits the plan before /to-spec. Reading only: /research.
incident-triageTriage an incident or outage from evidence: timeline, ranked causes, mitigations, one note. Read-only, never touches production; not log reads. Staging fixes: /staging-fix; local: /diagnosing-bugs.
integration-contractBuild a spec/PRD's cross-project contract after /to-tickets, or run its smoke gate before spec ship/PM handoff. A single-project consumer sweep may confirm no contract needed.
local-to-stagingPromote every project from origin/local to origin/staging: open each PR, merge on green, watch the Actions runs. Never pushes or goes past staging. Production is /staging-to-production.
orchestrate-herdrDispatch spec (PRD) sub-issues to worker tabs; monitor to done. Tab, CLI, or runtime context selects herdr; without any of these: /ask-kit. Tab status or actions with no dispatch: herdr companion.
orchestrate-t3Fan a spec (PRD) or issue list out to T3 Code worker threads, one worktree per group, and monitor them to done. Needs the T3 Code MCP. Inside herdr use /orchestrate-herdr; one issue is /tdd-loop.
pixel-auditAudit ONE page or route for visual conformance against Figma or reference screens, fix gaps with the UI library, and verify at element level. Not feature porting; never auto-chains.
plain-languageChat and prose cleanup at session start: fifth-grade English with bold glossary terms. Clean supplied prose; preserve mandated text. Not code, comments, ADRs, or specs.
polish-batchCreate a cosmetic QA punch list, with or without a ticket; capture, dispatch or verify rows. Figma matching: /pixel-audit. Behavioral bug reports or QA issue-filing sessions: /to-tickets.
port-featureUse only to port, migrate, or rebuild a complete feature from a reference implementation into a target stack. Writes one gap map, suggests /grill-with-field, stops. Never implements or auto-chains.
pr-feedbackAddress reviewer comments on an open GitHub PR: triage them for approval, apply accepted fixes, ship via /commit-push-pr, reply with commit SHAs. Reviewing a PR yourself is /code-review.
qa-escapeRecord a bug human QA found after delivery, including factory QA_RETURNED: reproduce it and identify the missed check. No prior delivery: /diagnosing-bugs.
release-notesWrite PM-friendly release notes, changelogs, or session summaries from git commits or the current session, in plain language. A PR summary for review is /code-review; an agent handoff is /handoff.
report-issueFile single or batch problem reports and screenshots as tracker issues. Code-traced, routed, labelled, prioritised, scheduled. Delivered-work bug: /triage. Fix now: /diagnosing-bugs.
risk-reviewAdversarial local review with a judge after implementation; PR merge-risk classification on request. Not open-ended impact questions. Standards/spec: /code-review. Comments: /pr-feedback.
staging-fixFix a staging bug without touching a server: fix locally with a test, then a PR to the staging branch with auto-merge. Use for staging is broken. Inspecting staging alone is not this skill.
staging-to-productionRead-only check of whether staging is ready for production in every project, with the owner's commands. Never opens, merges, pushes, or touches production. Local to staging: /local-to-staging.
tdd-loopKit TDD/red-green loop for changes, safe refactors and untested legacy code. Urgent production bug reports get local test-first fixes even if tests are requested later. Bare builds: /implement.
using-git-worktreesSet up or reuse a Git worktree when the user asks to work in a worktree, before the task skill runs. Not for branch-only requests, generic implementation, worktree questions, or cleanup alone.
writing-kit-skillsAuthor/edit a skill only when the request names agent-skills-kit or a kit skill, or explain its one-liners and word budget. Other skill work: skill-creator. Routing evals alone: maintainer task.
Build the work: a test-first loop, worktree setup, and fan-out of a spec's issues to herdr tabs or T3 Code threads.
31 skills
Understand existing behaviour and shape the work before building: read-only discovery, feature prompts, port gap maps, and cross-project contracts.
31 skills
Factory workflow skills: a conductor that reads where work stands and names the next gate, plus CI repair, agentic QA, risk-gated review, staging deploy watch, QA-escape intake, QA issue reporting, and incident triage.
31 skills
Move work through staging: promote local to staging, fix a staging bug through a PR, and check readiness for the owner's production promotion.
31 skills
Build and verify UI against a design source: design-system setup, per-page pixel audit, and the cosmetic polish tail.
31 skills
Set up a workspace and find your way around the kit: generated agent instructions, a skill router, and the kit's own authoring style.
31 skills
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.