本文へ移動
cccskills
無料GitHub で公開

provider-access-expand

Discovery batch that grows the Provider Access Launch queue. Use when asked to find more user-connected OAuth, API, API-key, remote MCP, marketplace, or partner-access campaigns for a capability, provider family, MCP/plugin catalog, or competitive gap. NOT for client Local Listings distribution.

インストール方法を見る

含まれるファイル(2)

  • SKILL.md5.2 KB
  • agents/openai.yaml254 B

SKILL.md(原文)

インストールする前に、エージェントに与えられる指示の中身を確認できます。

<!-- SYNCED COPY — do not edit here. Canonical: common-docs/skills/provider-access-expand/SKILL.md This file is distributed to every consuming repo by common-docs/meta/scripts/sync_skills.py. Edit the canonical, run the sync, and commit each repo. Edits made here are overwritten and lost. -->

Expand Provider Access

Add a researched, deduplicated batch to the normal AI Matrx Provider Access Launch Project. This skill discovers work; it does not create a parallel spreadsheet, register external accounts, request permissions, or claim that a provider is easy before scouting.

Keep Local Listings separate. This queue is for integrations a user connects to AI Matrx. Listing claims, location publishers, and data aggregators belong to common-docs/systems/marketing/local-listings/PLAN.md and its own project. Never add Data Axle, Localeze, Foursquare listing distribution, Yelp Listing Management, Apple business listings, or Google Business Profile here. If a listings-only seed already exists, mark it superseded and point to the Local Listings work; preserve its history.

Defaults

  • Add 15–25 campaigns for a named vertical or capability class.
  • Use the AI Matrx tenant organization 5dc930e9-bd65-44a1-8369-af773f6e1a5b.
  • One campaign is one independently approvable access boundary, not necessarily one provider.
  • Stable key: lowercase provider.capability.
  • Title: P0 | Provider | Capability.
  • New tasks start canonical inbox with Phase: discovered and no due check until scouting starts. The Projects JSON importer may store legacy incomplete; the product normalizes it to inbox.

Workflow

  1. Read common-docs/systems/integrations/provider-access/FEATURE.md when the workspace contains it.
  2. Open the live Provider Access Launch Project and inventory all existing campaign keys.
  3. Search the platform inventory before the public web:
    • installed/recommended plugins;
    • live tool.mcp_server and related tool.definition rows;
    • users.integration_connections and tool.mcp_user_conn;
    • provider code and integration UI;
    • Unified Credential Vault definitions, never secret values;
    • CRM provider parties/interactions;
    • common-docs/projects/* and existing task evidence.
  4. Use current official provider sources to confirm that each proposed capability or developer program exists. Do not turn search-result snippets, marketplace SEO pages, or memory into facts.
  5. Split providers where scopes, reviewer stories, production activation, or owners differ. Merge aliases that lead to the same authorization and implemented outcome.
  6. Assign priority:
    • P0: launch parity, existing MCP/plugin, social, enterprise, marketing/ads, core cloud, productivity, identity, or payments.
    • P1: major legal, medical, education, CRM, commerce, communications, or observability value.
    • P2: useful long-tail advantage.
  7. Create each campaign through the normal Projects/Tasks product path. If an exact key exists, enrich the existing task; never create a duplicate.
  8. End with counts by priority/category, the campaigns added, aliases collapsed, and any candidate rejected because no official/current route could be established.

Minimum new-task card

Campaign: provider.capability
Phase: discovered
Category: <category>
Access path: <oauth_review|oauth_self_service|api_key|remote_mcp|partner|marketplace|research>
Easy candidate: <yes|no|research> (unverified until scouting)
Existing asset: <link/identity or none found>
Source seed: <official URL or official sources required during scouting>
Next action: Run $provider-access-scout for this campaign.
Next check: none until scouting completes

Do not put secrets, copied access tokens, guessed requirements, or a submission recommendation in a new discovery task. “Easy candidate” is a scouting hypothesis only.

Deduplication decisions

  • Same provider + same authorization + same implemented outcome: one campaign.
  • Same provider + unrelated scope family or materially different review: separate campaigns.
  • Official remote MCP and a distinct first-party OAuth/API product: separate if they produce independently useful connections.
  • Developer sandbox and production activation: one campaign with separate gates unless either can be useful and verified independently; then split them.
  • Existing detailed plan: link it. Do not re-copy its history into the task.
  • Existing connected integration: create or update the task only if verification/follow-through is incomplete; otherwise mark it verified with evidence.

Example requests

  • “Use $provider-access-expand to add legal-practice and e-discovery providers.”
  • “Use $provider-access-expand to reconcile every recommended plugin against the launch queue.”
  • “Use $provider-access-expand to find another 20 education access campaigns.”

レビュー

まだレビューはありません。使ってみた感想をお寄せください。

同じリポジトリのスキル

概要と使いどころ

Admin debug system wiring for the floating AdminIndicator's live debug data and Copy Full Context. Use when adding debug visibility to a route or feature, wiring useDebugContext or the debug panel, capturing console errors, or enabling the copy-context workflow for a page.

日本語の概要は準備中です。原文の説明を表示しています。

armanisadeghi/ai-matrx32026年10月13日 更新

Compact two-icon Copy / Copy-for-AI controls (components/agent-copy). Use when adding copy buttons to a row, card, list, or record; merging duplicate Copy/JSON/AI controls; continuing the copy rollout; or writing a Copy-for-AI payload. NOT for markdown content actions (use rich-document-actions).

日本語の概要は準備中です。原文の説明を表示しています。

armanisadeghi/ai-matrx32026年10月13日 更新

Disclosing a surface's existing fixed AI jobs in the shell's top Agents menu. Use when a page, panel, overlay, or window already runs a mandate behind a button, assist, automatic action, or mode; when the agent-disclosure guard names a file; or during a surface check.

日本語の概要は準備中です。原文の説明を表示しています。

armanisadeghi/ai-matrx32026年10月13日 更新

The watch-fix-rerun method for making an existing platform agent or automated process efficient and correct: baseline its ledger, run one unit yourself, fix the class behind every wasted call, rerun, record. Use when asked to improve, optimize, watch, or 'make efficient' an agent, a sandbox session, a sync job, or any recurring automated process.

日本語の概要は準備中です。原文の説明を表示しています。

armanisadeghi/ai-matrx32026年10月13日 更新

Redux state for agent execution and firing agent shortcuts. Use when editing features/agents/redux/, building agent UI, creating a conversation, touching assembleRequest or NDJSON stream state, wiring agent-state selectors, adding a per-conversation capability, or triggering a shortcut from a button, menu, or code (useShortcutTrigger, launchAgentExecution).

日本語の概要は準備中です。原文の説明を表示しています。

armanisadeghi/ai-matrx32026年10月13日 更新

The provision for a call site: the exhaustive menu of values that place in the code can realistically produce. Use before creating or fixing a mandate or agent that reasons about more than its own input, when deciding whether an agent can answer at all with what it is sent, or when its output 'looks right' but nobody checked it could be. NOT for building the agent (use create-agent).

日本語の概要は準備中です。原文の説明を表示しています。

armanisadeghi/ai-matrx32026年10月13日 更新

armanisadeghi のスキルをすべて見る

このスキルの問題を報告する