AgentSkillsCN

tensorpm-agentic-pm

由 TensorPM 提供支持的代理式项目管理。通过 MCP 工具与 A2A 协议，高效管理项目、行动事项与工作空间。面向代理的上下文驱动型 AI 项目管理。

SKILL.md
--- frontmatter
name: tensorpm-agentic-pm
description: 'Agentic project management powered by TensorPM. Manage projects, action items, and workspaces through MCP tools and A2A protocol. Context-driven AI project management for agents.'
compatibility: Requires the TensorPM desktop app to be running for MCP tool access and A2A communication. Available on macOS, Windows, and Linux.

TensorPM Skill

Use this skill for AI-powered, context-driven project management inside a running TensorPM desktop app. TensorPM itself is free. For AI capabilities outside MCP (A2A), use your own API key (BYOK) or create an account.

When To Use

  • •You need to list, create, or update TensorPM projects or action items.
  • •You need to switch/list workspaces.
  • •You need to set AI provider keys through TensorPM (set_api_key).
  • •You need conversational project-level changes via A2A (message/send).

When Not To Use

  • •The request is only about website/account/billing pages.

Installation (Agent CLI)

Use one of these agent-friendly CLI install methods:

bash
# macOS
brew install --cask neo552/tensorpm/tensorpm
powershell
# Windows (PowerShell)
winget install --id Neo552.TensorPM --exact --accept-package-agreements --accept-source-agreements
bash
# macOS / Linux fallback installer script
curl -fsSL https://raw.githubusercontent.com/Neo552/TensorPM/main/scripts/install.sh | bash
powershell
# Windows fallback installer script
irm https://raw.githubusercontent.com/Neo552/TensorPM/main/scripts/install.ps1 | iex

Runtime Prerequisites

  1. •Start TensorPM desktop app.
  2. •For MCP usage with external AI clients: ensure client integration is installed once (via Settings -> Integrations or A2A POST /integrations/mcp/install).
  3. •For A2A usage: verify local endpoint http://localhost:37850.

MCP vs A2A Routing

TaskUseWhy
Structured action-item CRUDMCP toolsDirect typed operations
Set provider API keysMCP set_api_keyDedicated secure write-only tool
Project-wide/contextual changesA2A message/sendManaged by project manager agent
HTTP-based automation/client integrationA2A REST/JSON-RPCEndpoint-first integration path
Multi-turn planning with conversation stateA2A with contextIdNative conversation continuity

Rule of thumb:

  • •Prefer MCP for explicit CRUD operations.
  • •Prefer A2A for high-level intent and context-aware planning.

Minimal Workflow

  1. •Verify TensorPM is running.
  2. •Choose MCP vs A2A via the routing table above.
  3. •Execute operation.
  4. •Read back result (list_*, get_project, or A2A read endpoint) to confirm state.
  5. •Summarize applied changes and any follow-up action.

References

Notes