skills.rest
aaronjmars
aaronjmarsCOMMUNITY
Explore all skills by this author →
𝕏@aaronjmarsGitHub

executor-mcp

Execute tasks through an MCP proxy catalog of connected integrations with per-tool policies.

714|255|Updated Mar 4, 2026
Search Tags:#mcp#Automation#api proxy#tool catalog#Policy enforcement#oAuth#Integrations
One-click install
npx skills add https://github.com/aaronjmars/aeon --skill executor-mcp
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: executor-mcp
Source: https://github.com/aaronjmars/aeon/tree/main/skills/executor-mcp
Command: npx skills add https://github.com/aaronjmars/aeon --skill executor-mcp

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

It lets an agent run a single task against every integration an operator has connected (MCP servers, OpenAPI specs, GraphQL APIs) through one Executor Cloud MCP endpoint, without ever handling credentials, while enforcing per-tool allow, approval, and block policies.

Core Features & Use Cases

  • Unified tool catalog discovery: Enumerates the mcp__executor__* tools exposed by the Executor server each run and maps the task onto them, never assuming an integration exists.
  • Policy-governed execution: Honors allow, require-approval, and block outcomes per call, caps runs at 15 calls, and sequences irreversible writes as the final action.
  • Structured reporting: Delivers exactly one Markdown notify per run and appends a result-coded log entry (EXEC_OK, EXEC_AUTH_STALE, EXEC_POLICY_BLOCKED, etc.) to the daily memory log.
  • Use Case: Ask the agent to list open Linear issues and summarize them by project; it discovers the Linear integration through Executor, calls the tools within policy, and reports the summary plus any pending approvals.

Quick Start

Run this skill with a task such as "list my open Linear issues and summarize by project" after connecting Executor in the dashboard MCP panel.

Frequently Asked Questions about executor-mcp

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I run a task across connected integrations through one MCP endpoint?▼

Pass the task as the skill's input and it discovers the Executor server's exposed tools, maps the task onto them, and executes with the fewest calls possible. The Executor Cloud endpoint proxies every connected MCP server, OpenAPI spec, and GraphQL API as a single catalog.

How does Executor MCP handle credentials for upstream integrations?▼

Credentials live in Executor Cloud and are attached upstream per call, so the agent never sees them. The server connects via OAuth from the dashboard MCP panel, with tokens refreshed each run by the mcp-oauth-refresh script.

What happens when a tool call requires approval or is blocked?▼

A call requiring approval parks until a human approves it in the Executor console; the run finishes what allowed tools can do and reports EXEC_APPROVAL_PENDING. Blocked calls are reported as EXEC_POLICY_BLOCKED and never worked around.

Why does the skill report EXEC_NOT_CONNECTED or EXEC_AUTH_STALE?▼

EXEC_NOT_CONNECTED means no mcp__executor__* tools are callable, so the server is not connected or its secrets are missing. EXEC_AUTH_STALE means calls return 401 because the OAuth refresh failed, requiring a re-connect in the dashboard.

What are the limitations of running tasks through the Executor proxy?▼

Each run handles one task with a maximum of 15 tool calls because upstreams are rate-limited and metered. If a needed integration is not in the catalog, the run exits with EXEC_NO_INTEGRATION rather than substituting another tool.

Related Skills

Explore More Software Engineering Skills→
LumenYoungLumenYoung

mcp2cli

Convert MCP endpoints, OpenAPI specs, and GraphQL schemas into executable command-line interfaces.

Community
Advanced
swarm-ai-researchswarm-ai-research

Janitor

Delete stale .notify-* files, .pending-notify-temp/ artifacts, and .outputs/ chain artifacts by TTL.

Official
Intermediate
bashartechbashartech

linkedin-automation

Automate LinkedIn post drafting, approval, and publishing with human review.

Community
Intermediate
rsdouglasrsdouglas
30

janee

Inject credentials into proxied API requests with policy checks and audit logs.

Community
Advanced