agent-reach

$npx mdskill add JasonxzWen/harness-hub/agent-reach

Route read-only internet research through an installed Agent Reach CLI.

  • Solves the need for read-only web research without installing new tools.
  • Depends on an already installed Agent Reach CLI and its active backends.
  • Runs agent-reach doctor --json to select only active backends for the channel.
  • Reports missing prerequisites or delegates to narrower capabilities when available.

SKILL.md

.github/skills/agent-reachView on GitHub ↗
---
name: agent-reach
description: Load when a task needs read-only internet research or content retrieval through an already installed Agent Reach CLI; diagnose available backends first, never install, upgrade, authenticate, configure, or copy credentials on the user's behalf.
license: MIT
metadata:
  source: https://github.com/Panniantong/Agent-Reach
  revision: e825f6740d24c6c315c3b0dc41907e6c87ff39a5
---

# Agent Reach

Use Agent Reach only as a router to read-only internet backends that already
exist in the user's environment.

## Availability gate

1. Tell the user that Agent Reach is being checked.
2. Run agent-reach doctor --json.
3. Use only a backend that the returned evidence marks active for the requested
   channel.
4. If the CLI or required backend is unavailable, report the missing user-level
   prerequisite. Do not claim that migration installed or configured it.
5. When another installed project or Host capability already owns the source,
   prefer that narrower capability.

Before invoking an active backend, inspect its local --help output when the
exact read-only command is not already proven by the doctor result. Do not guess
commands, configuration keys, or login state.

## Hard boundaries

- Read and search only. Do not post, comment, like, follow, submit forms, open
  pull requests, or perform any other remote write.
- Do not run Agent Reach install, update, uninstall, configure, login, or
  check-update commands.
- Do not install or upgrade Python, Node.js, GitHub CLI, mcporter, browser
  tooling, platform CLIs, system packages, or MCP servers.
- Do not write ~/.agent-reach, Host global configuration, shell profiles, or
  credential stores.
- Do not copy, export, request, or persist cookies, tokens, browser profiles, QR
  sessions, or other login state.
- Do not weaken sandbox, approval, or repository authorization boundaries.

If an unavailable channel is still required, stop at a concise prerequisite
report: missing component, why it is required, and the user-level action the
user may choose to take outside Harness Hub. Continue with an already available
read-only source only when it can honestly satisfy the request.

## Evidence

In the final answer, distinguish:

- sources actually retrieved;
- channels unavailable or skipped;
- authentication or dependency status reported by doctor;
- any fallback source used instead.

Never infer availability from this Skill being present.

More from JasonxzWen/harness-hub

SkillDescription
agent-interaction-auditLoad when the user asks for private agent interaction audits from local Codex/Claude Code traces, task profiles, tool-call decisions, collaboration bottlenecks, prompt/rule or automation-log audits, closeout retrospectives, or evidence-backed Skill/project-rule/Eval/SOP/OKF improvements.
animation-vocabularyLoad when a user describes a Web animation or motion effect without knowing its name and needs the precise term or close alternatives; use only for naming, not design, implementation, or review.
apple-designLoad when a frontend task needs Apple-platform design guidance or physical, gesture-driven, velocity-aware, interruptible interaction; skip generic visual styling and use review-animations for report-only motion review.
beautiful-feishu-whiteboardLoad when creating or restyling a polished editable Feishu or Lark whiteboard, infographic, diagram, poster, or visual explainer; do not load for generic SVGs, static images, or other whiteboard platforms.
clone-websiteLoad when the user explicitly asks to clone, reverse-engineer, or rebuild a website they own or are authorized to reproduce; do not load for generic redesign, inspiration, scraping, phishing, impersonation, or copying third-party brands.
codebase-to-courseLoad when a user wants a source-backed interactive HTML course that explains how a codebase works; do not load for chat tutoring, repository review, articles, slide decks, or general browser artifacts.
context-guidanceFixture skill used to represent an additional standard skill context surface.
decision-uiLoad when Codex 即将要求用户在 2-3 个互斥方案中做真实决定、准备询问阻塞性澄清问题或需要确认高影响取舍;将选择转换为宿主原生可点击 UI,并在工具不可用时诚实文本回退。不要用于低风险实现细节、只有一个合理答案的问题、普通信息采集或不需要用户决策的完成汇报。
design-taste-frontendLoad when a task needs anti-template frontend visual direction, design-read calibration, or pre-flight critique for landing pages, portfolios, marketing pages, or redesigns; do not load for dashboards, data tables, multi-step product UI, routine frontend logic, HTML reports, slide decks, or generic code explanation.
effective-interactLoad when agent pauses to report relatively complex information needing Chinese-first clear complex communication, alignment, multi-option choice, status/incident, long-task fact ledgers, implementation plans, reviews, maps, explainers, evidence, risks, validation, handoff; choose plain text/Markdown/visual Markdown/HTML by decision cost; do not load merely because answer is long; skip trivial chat/bundled apps.