Agent skills published by Forward Path for use with Claude Code and other agents that follow the skills.sh format.
Install any skill in this repo with the skills CLI:
npx skills add forwardpathai/skillsOr install a single skill by name:
npx skills add forwardpathai/skills/issue-writer
npx skills add forwardpathai/skills/open-pr
npx skills add forwardpathai/skills/release
npx skills add forwardpathai/skills/stack-pr
npx skills add forwardpathai/skills/cloud-agent-pr-stats
npx skills add forwardpathai/skills/cloud-agent-triage
npx skills add forwardpathai/skills/backlog-hygiene
npx skills add forwardpathai/skills/ticket-refiner
npx skills add forwardpathai/skills/azure-infra-setup
npx skills add forwardpathai/skills/engineering-projects-on-track
npx skills add forwardpathai/skills/mobile-ui
npx skills add forwardpathai/skills/mobile-ui-implement
npx skills add forwardpathai/skills/web-ui
npx skills add forwardpathai/skills/customer-deployment-package
npx skills add forwardpathai/skills/review-mp4
npx skills add forwardpathai/skills/azure-hibernate
npx skills add forwardpathai/skills/qa-test-plan
npx skills add forwardpathai/skills/qa-test-plan-setup
npx skills add forwardpathai/skills/teach-web-actions
npx skills add forwardpathai/skills/product-foundation
npx skills add forwardpathai/skills/linear-release-setup
npx skills add forwardpathai/skills/linear-backlog-grill
npx skills add forwardpathai/skills/plan-refiner
npx skills add forwardpathai/skills/plan-context-imager
npx skills add forwardpathai/skills/speclang-writer
npx skills add forwardpathai/skills/poc-from-sow
npx skills add forwardpathai/skills/poc-to-product-architecture
npx skills add forwardpathai/skills/architecture-to-linear-plan
npx skills add forwardpathai/skills/standup-report
npx skills add forwardpathai/skills/inbox-brief
npx skills add forwardpathai/skills/project-update-writer
npx skills add forwardpathai/skills/eng-report-effort-estimate
npx skills add forwardpathai/skills/setup-preview-env| Skill | Description |
|---|---|
| issue-writer | Write Linear issues an AI coding agent can execute without follow-up questions. |
| open-pr | Open a GitHub PR from local changes via a Linear issue, Linear's git branch, commit, and gh. |
| release | Cut a production GitHub release from the default branch — semver tag, release notes, breaking-change gate, and CI watch. |
| stack-pr | Open a stacked/dependent GitHub PR based on the current branch (sibling of open-pr) via a Linear issue, commit, and gh. |
| cloud-agent-pr-stats | Count PRs opened by cloud coding agents (Cursor, Codex, Devin, Claude, Copilot) across a GitHub org, by branch-name prefix. |
| cloud-agent-triage | Evaluate whether Linear tickets suit autonomous Cursor cloud agents (why yes / why not), triage a repo's backlog honoring blockers, and optionally delegate the ready ones to Cursor. |
| backlog-hygiene | Scan a Linear project's backlog for relevance — flag stale, already-shipped, or duplicated issues using update age, code/PR evidence, and ticket similarity — then apply confirmed actions one at a time. |
| ticket-refiner | Interview the person who knows why a Linear ticket exists and rewrite it to pass issue-writer's agent-ready bar; the gap-driven interview step neither backlog-hygiene nor cloud-agent-triage has. |
| engineering-projects-on-track | On-track % for ForwardPath Engineering Linear projects in In Progress or UAT, from the latest project status update. |
| azure-infra-setup | Author Forward Path Azure infrastructure (Terraform/Bicep) — shared ACR, OIDC/RBAC, environments, Key Vault, Container Apps. |
| mobile-ui | Turn a SOW into a grounded Expo SDK 54 screen spec and premium portrait UI mockups (via Google Gemini or OpenRouter, with a consent-gated on-device Apple-Silicon fallback when no cloud key is available). |
| mobile-ui-implement | Turn mobile UI mockups (PNG screens or a PDF deck) into per-screen implementation blueprints, then build them in an Expo Go-compatible app — decomposing each screen into layers, components, states, and animations a weaker model can build one at a time. |
| web-ui | Turn a SOW and/or a real Next.js + Tailwind codebase into a grounded web screen spec and consistent desktop mockups (via Google Gemini or OpenRouter, with a consent-gated on-device Apple-Silicon fallback when no cloud key is available) — anchored by a brand style board and a persistent app-shell (nav/sidebar) so every screen looks like one credible product. |
| customer-deployment-package | Build a customer-facing deployment handoff — external Terraform/Bicep, setup instructions filled into the Notion template and saved to the deployments DB, then exported and zipped with a Windows-safe name. |
| review-mp4 | Understand an mp4 (local or URL): extract frames with ffmpeg, pick the sharpest in-focus frame per window via variance-of-Laplacian blur detection (Python or Node), then read them to answer questions. |
| azure-hibernate | Hibernate a live client project's Azure resource group to minimum cost (scale/stop App Service Plans, web apps, SQL, Redis, databases via the az CLI) and wake it for retesting — reversibly, recording state before every change. |
| qa-test-plan | Add QA test cases to the Notion QA Test Plan database — generate candidates from a feature change or the whole app, get QA approval, then append rows QA works through during full regression runs. |
| qa-test-plan-setup | Per-project one-time creation of a <Project> — QA Test Plan Notion database with the fixed Forward Path schema (test case, feature, steps, expected result, category, priority, status). |
| teach-web-actions | Learn a website by recording a user-driven Chrome session (HAR + UI steps) via Playwright codegen, distill it into a reusable lesson (endpoints, payloads, parameter knobs, auth), then replay a variation with new parameters as an API call or as UI navigation captured to mp4. |
| product-foundation | The standard stack and conventions for building Forward Path products — Bun monorepo, Next.js App Router, Hono RPC APIs, Drizzle/Postgres, TanStack Query, Better Auth, Azure. Scaffold new repos, add modules, and enforce code conventions. |
| linear-release-setup | Set up GitHub Release workflows that publish containers, keep a scheduled Linear release in progress from main merges, and complete it when the customer GitHub Release ships. |
| linear-backlog-grill | Grill Linear project backlogs for execution readiness — grade tickets, interview for missing specification, split oversized work, and rewrite issues to pass the agent-ready bar. |
| plan-refiner | Harden an existing Cursor plan so a weaker executor model can implement it without judgment calls — hunt unknowns, resolve them via codebase/web research or a one-gap-at-a-time user interview, anticipate non-obvious scenarios, and rewrite the plan file in place as unambiguous, verifiable statements. |
| plan-context-imager | Gather the codebase context a plan depends on, render it into dense PNG pages via pxpipe's renderTextToImages, and embed the images plus a per-page index into the plan so the executor reads them instead of re-grepping the codebase each step. |
| speclang-writer | Turn a plan into a SpecLang specification — a structured, natural-language Markdown document (the single source of truth) that captures a system's behavior and its pinned implementation details so an AI toolchain or executor can generate the code from the spec. |
| poc-from-sow | Turn a SOW into a customer-validatable, continuation-ready POC codebase on the Forward Path stack — buildable, runnable, demo-deployable, with an agent-ready handoff pack — by orchestrating web-ui / mobile-ui / product-foundation / azure-infra-setup, then closing with a poc-to-product-architecture canvas + DOCX. |
| poc-to-product-architecture | Turn a SOW and POC repo into a production system-design canvas — architecture, gap audit, Azure resource map, Bicep skeleton, security/reliability, and cost estimate, constrained to the Forward Path stack and customer-deployable Bicep. |
| architecture-to-linear-plan | Turn the poc-to-product-architecture deliverable (canvas + DOCX) and any web-ui/mobile-ui mockups into a delivery plan — Linear milestones with proposed timelines and executable functional + design tickets — then, once approved, create the project, milestones, and issues in Linear via the Linear MCP. |
| standup-report | Generate a narratable daily standup (yesterday / today / blockers) for yourself or a named user from Linear issues, GitHub merged/open PRs, and local Cursor chat transcripts. |
| inbox-brief | Context brief of today's Outlook inbox (or named emails) — scrapes Outlook Web via a persistent Playwright profile (no Graph API/admin consent), filters GitHub/automated noise, and enriches each human thread with related Linear issues and local Cursor chats. |
| project-update-writer | Summarize everything shipped in a Linear project since its last status update — baseline from the latest update, issues completed since grouped by capability area — and post a new project update with health after the user confirms. |
| eng-report-effort-estimate | Generate the combined Engineering Report + Effort Estimate document in Notion for a custom build — understanding check with price levers, gated generation, fixed-price effort estimation with internal QA phase, and a post-generation coverage audit. |
| setup-preview-env | Set up ephemeral per-pull-request Azure preview environments for a containerized app — shared/per-PR Terraform, GitHub Actions deploy/teardown, auth redirects, and runbooks distilled from the ButtconRAG preview lane. |
Each skill lives in its own directory at the repo root with a SKILL.md entry point that has YAML frontmatter. A skill may also bundle reference docs and utility scripts alongside it:
skills/
└── <skill-name>/
├── SKILL.md # required entry point
├── REFERENCE.md # optional reference docs
└── scripts/ # optional utility scripts
Reference files and scripts must be addressed relative to the skill's own directory (never hardcode an absolute install path), so the skill keeps working wherever skills.sh installs it.
The name in frontmatter must match the directory name. The description is what agents match against to decide whether to invoke the skill — make it specific about when the skill should fire.
MIT