juling-qianjiang

Parse soul.md constraints and assign primary/support spirits with governance-aware authority modes.

10|1|Updated May 4, 2026
One-click install
npx skills add https://github.com/isLinXu/under-one --skill juling-qianjiang
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: juling-qianjiang
Source: https://github.com/isLinXu/under-one/tree/main/underone/skills/juling-qianjiang
Command: npx skills add https://github.com/isLinXu/under-one --skill juling-qianjiang

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) components.

What problem does it solve?

拘灵遣将 solves the problem of keeping multi-tool, multi-agent task execution reliable by selecting the best available tool/spirit, applying deterministic degradation (protect) or simulated/inner fallback (possess), and enforcing boundary-aware governance to reduce rebellion risk and prevent overreach.

Core Features & Use Cases

  • 多维度统御匹配:根据能力匹配(精确/别名)、健康度(成功率/延迟/错误率)、负载均衡与历史质量进行综合评分,输出可执行调度方案。
  • 灵魂边界与风险封控:从 soul.md/souls/ 解析 persona、capabilities、limits/taboo 与调用规则,计算 rebellion_risk,并在高风险时切换 authority_mode(sealed-command / cautious-attunement / direct-command)。
  • 降级保护与强制服灵:当工具不可用时在 protect/possess 两种策略下执行缓存/模拟/本地计算/内化替代,并生成可追踪的 recovery_plan 与 escalation_contract。
  • 阵型调度与协同语义:支持 single-possession / dual-attunement / night-parade,生成主副灵分工、调用顺序 spirit_queue、并行通道语义与执行检查点。

Use case examples:

  • 你有一组需要“搜索 + 代码分析 + 写作”的任务链,并且部分外部工具在运行时可能不可用或质量下降;该 Skill 能自动选择可用灵体/工具并在故障时按策略降级,同时输出治理与恢复契约,便于上层 agent 做守门决策。
  • 你希望子代理/工具在执行前读取 soul.md 限制,避免越权写入或越界行为;该 Skill 能把 limits / invocation rules 变成边界约束与调用规约,并在风险升高时收缩权限。

Quick Start

Use the juling-qianjiang skill to dispatch a task list from tasks.json with tool candidates from spirits.json in protect mode by running: python scripts/dispatcher.py tasks.json spirits.json protect.

Frequently Asked Questions about juling-qianjiang

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

FAQPage Schema
How does multi-agent load balancing and tool orchestration handle risk assessment?

Multi-agent load balancing uses risk assessment by scoring capability match, health, load, and historical quality, then parsing soul.md constraints to calculate rebellion risk and switch authority modes. This prevents boundary violations during multi-tool task dispatch.

How do I prevent boundary violations and overreach when dispatching tasks to multiple agents?

To prevent boundary violations, you can parse soul.md constraints to extract persona capabilities, limits, and invocation rules. The system enforces boundary-aware governance by assigning primary and support spirits with governance-aware authority modes.

What is the best way to implement fallback protection for unavailable tools in an agent ops workflow?

The best way to implement fallback protection is executing deterministic degradation or simulated inner fallback strategies. When tools are unavailable, the system uses caching, local computation, or internal alternatives, generating a tracked recovery plan and escalation contract.

Does multi-agent formation support single, dual, and parallel execution with fallback recovery?

Yes, multi-agent formation supports single-possession, dual-attunement, and night-parade modes. It generates primary and support spirit divisions, invocation queues, parallel channel semantics, and execution checkpoints for batch dispatch and fallback recovery.

Can I use soul.md parsing to constrain agent permissions and avoid unauthorized writes?

Yes, you can use soul.md parsing to convert limits and invocation rules into boundary constraints. When calculated rebellion risk rises, the system automatically shrinks permissions by switching to sealed-command or cautious-attunement authority modes to prevent overreach.

When should I not use automated spirit dispatch for multi-tool execution?

You should avoid automated spirit dispatch when your tasks require deterministic external tool availability without degradation, as the system relies on simulated fallback and local computation alternatives when health scores drop or tools fail during execution.