OPC Skills
Language: 中文说明 · Full English docs
Current version: v0.6.2
opc-skills is a Codex Skill for shipping consumer-facing apps together with their B-side operator consoles. This hub is the short map. Full detail lives in README.en.md and README.zh-CN.md. Agent rules are defined in SKILL.md; do not treat this page as a substitute.
Install
One-liner (creates ~/.codex/skills if needed):
npx opc-skills install
curl -fsSL https://raw.githubusercontent.com/BiuBiuHu/opc-skills/main/scripts/install.sh | bash
npx opc-skills update uses a verified Git source cache and synchronizes the installed skill snapshots. Existing .agents, Cursor and Claude skill surfaces are included; per-install local edits use a verified baseline and three-way merge. Any conflict leaves all installations unchanged, with backups retained. doctor checks provenance, versions and local overrides rather than requiring .git in the installed directory. Git and Python 3.9+ are required; see local installation for adoption, isolated tests and current CLI availability.
Optional Codex Agent Plugin install (codex plugin marketplace add BiuBiuHu/opc-skills, then codex plugin add opc-skills@opc-skills) is documented in docs/skills-hub.md, together with $skill-installer prompts. openai/skills curated listing is deprecated; this repo is not in the public Plugins Directory.
Manual clone:
mkdir -p ~/.codex/skills
git clone https://github.com/BiuBiuHu/opc-skills.git ~/.codex/skills/opc-skills
Or clone elsewhere and symlink into ~/.codex/skills/opc-skills.
Deployment platforms
Vercel and Cloudflare are supported. See Cloudflare deployment for Pages, Workers Static Assets, routing, and migration. For APIs, auth, email, or AI services, native Workers migration covers request-driven execution, existing PostgreSQL/Hyperdrive, streaming cleanup, CI, real acceptance, and measured cost. Containers require a reason; moving applications does not imply moving the database. Hosting choice follows the user and project configuration.
Trigger phrases
In Codex, use one of these intents:
Start OPC Skills
Start the consumer app delivery workflow
Run the full consumer app + operator console loop
Use opc-skills to analyze this mobile requirement
Use opc-skills to design this consumer feature
Use opc-skills to design the matching ops console
Generate Chinese and English tech article drafts from this release
Chinese trigger phrases are listed in README.zh-CN.md.
The main agent then reads SKILL.md and enters requirement validation, UI, architecture, implementation, testing, integration, release, or post-release-content.
What it is
A reusable Codex Skill that walks the main agent through a reviewable delivery loop:
clarify → PRD → UI → architecture → implement → test → integrate → release → optional post-release content
It covers user-side App / Web / mini-program / desktop work, the matching ops, review, config, and support consoles, and the primary API / auth / worker path behind them. The skill keeps reusable methods, gates, templates, and role placeholders. Project names, accounts, domains, service aliases, and deployments stay in the project's own fact sources.
Default planning mode is manual-gated; wait when implementation is not yet authorized or the user explicitly requires a plan review. Reuse valid authorization for the same scope instead of asking again after writing documents.
Who it's for
- Solo builders and one-person companies (OPC) who need a complete path from a pain point to a shippable release
- Small teams that ship a consumer product and the operator console that keeps it running
- Work that needs evidence, not only a green build: real pages, real clients, staging, test reports, and production regression
What you get
- Evidence-gated workflow: real pages and clients, staging, test report, code review, release-plan sign-off, production smoke, production-domain regression, then an observation window
- Templates under
templates/ for requirement validation, PRD, UI prototype, architecture, change impact, implementation plan, test strategy, release plan, evidence manifest, ops runbook, and post-release content
- Role files under
agents/ (Explorer, Frontend, Backend, QA, Integration, DevOps, Content). The skill does not call subagents itself; the main agent uses runtime tools only when the user explicitly allows it
- Release-integrity reference and
scripts/check_release_integrity.sh for candidate commit, deployment commit, default-branch merge, and worktree reconciliation
- Chinese-default PRs: new Code / Release / docs PR titles, bodies, and checklists default to Chinese unless repo policy or the user says otherwise
- Optional bilingual post-release articles via Content Agent (
suggest / draft / publish). Page release is never blocked by article work, and articles are not published silently
- Feature docs written to
docs/<FEATURE_NAME>/; helpers include scripts/start_rnd.sh, scripts/check_project_coupling.sh, and scripts/prepare_post_release_content.sh
Workflow snapshot
goal clarification
-> requirement validation / PRD
-> consumer UI + B-side console prototype
-> architecture + call graph + change impact
-> implementation plan
-> local mock / client automation
-> staging + real API
-> test report + code review
-> release-plan review + related-system sign-off
-> default-branch merge freeze
-> production + smoke + domain regression
-> worktree / integrity closeout
-> optional bilingual post-release content
This is the map only. The language pages and SKILL.md own the gates.
Further reading
Related notes from One PC:
Docs
License
MIT
