Borrowing it
Nothing to install: this file belongs to Utopia-V/mixagents. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/Utopia-V/mixagents/main/.agents/skills/maintain-mixagents/SKILL.mdgit clone --depth 1 https://github.com/Utopia-V/mixagentsWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/utopia-v/mixagents/maintain-mixagents)<a href="https://agentmods.dev/skills/utopia-v/mixagents/maintain-mixagents"><img src="https://agentmods.dev/badge/skills/utopia-v/mixagents/maintain-mixagents.svg" alt="Measured on agentmods" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00058 | $0.01841 |
| Opus 5 | $0.00029 | $0.00920 |
| Sonnet 5 | $0.00012 | $0.00368 |
| Haiku 4.5 | $0.00006 | $0.00184 |
Grade C, and why
maintain-mixagents scanned grade C with 1 finding against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured yesterday.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Harvests environment variableshighData exfiltration
Enumerating or grepping the environment for keys collects credentials unrelated to what the mod says it does.
decline explicitly; never collect credentials or expand route authority. How it starts
The opening of the file, as written. The whole thing — 158 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Maintain Mixagents
The repository contains two independent integrations and the provider-neutral Broker project. Select the affected package first and keep its lifecycle, distribution surface, and evidence separate from the others. Do not introduce a shared abstraction merely because packages involve agents, providers, or similar transports.
MixAgents Broker
Read the Broker product and status owner, its accepted backend and lifecycle policy, its accepted controller contract, its local project instructions, and its security owner before changing routing, worker lifecycle, provider data, credentials, permissions, tools, or paid-call behavior.
Broker exists to schedule workers across model providers without globally switching the controller's provider. Preserve that parent outcome rather than reducing the package to a DeepSeek-specific adapter or a one-shot model tool. Keep these boundaries explicit:
- Name the effective controller, provider, model, backend, worker identity, and lifecycle owner truthfully.
- A backend owns only the lifecycle semantics it can actually provide. Never present an externally managed worker as a native Codex child.
- Provider selection does not grant tool, filesystem, network, or mutation authority. The effective harness and environment own those permissions and side effects.
- Missing capability, credentials, or transport fails closed. Never silently change provider, model, backend, or task carrier.
- Prefer native Codex children only when the full cross-provider capability contract passes; otherwise select the qualified App Server fallback before dispatch. Never migrate a live Agent as an implicit retry.
- Route from a versioned qualification record. Do not trial-spawn a native child per Agent; requalify only when the Codex build, plugin, route, credential identity, environment, permissions, or relevant protocol changes. Start with offline checks, and require explicit authorization for a live canary.
- Promote a fully evidenced qualification atomically for new Agents only. Demote the affected key on a provider, transport, identity, or lifecycle contract violation, not an ordinary task failure. Reuse the same creation through App Server only after proving native accepted no child and began no external effect; ambiguity must not create a second worker.
- Preserve the native-sized Agent thread contract:
routes,spawn_agent,send,wait_agent,interrupt_agent, andlist_agents; a small status vocabulary; and final text or an error. Do not reintroduce a generic Job, revision, pending-action, cancellation, artifact, or release protocol without a demonstrated consumer that App Server threads cannot serve. - Completion, failure, and interruption end the current turn, not the Agent thread. Follow-up input may start another turn on the same App Server thread.
- App Server owns managed thread history, turn results, and recovery. Persist only non-secret runtime metadata that cannot be reconstructed from it.
- Scope runtime processes by route and effective access. The
$brokerSkill routes qualified native Agents through Codex collaboration tools directly; the local STDIO MCP server owns App Server-backed Agents. - MCP tools accept only preconfigured routes, never inline credentials, endpoints, headers, or provider definitions.
- Send only the complete assignment and context explicitly selected for the worker. Managed workers use the supplied checkout directly; the controller serializes overlapping writers or prepares a Codex worktree before dispatch.
- Relay transient non-secret approval or user-input requests through the host only when it advertises elicitation during an associated call. Otherwise decline explicitly; never collect credentials or expand route authority.
- Treat App Server
turn/interruptas a local turn result. State the provider compute/billing limitation instead of inventing a remote-cancellation claim. - Treat probes as disposable evidence; record only decision-grade results in their durable project owner.
What ships with it
1 file beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- yesterday First seen · 158 lines · 58 tokens per session scan C 7d336600e790
maintain-mixagents is a skill published in the GitHub repository Utopia-V/mixagents (185 stars, last pushed 2d ago), licensed MIT. It adds 58 tokens to every session and 1,841 once invoked, about $0.0003 per session on Opus 5. A static security scan graded it C with 1 finding (harvests environment variables). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-09-05.
Other skills, from other repositories
shortfilm-prompt
电影感 AI 视频提示词的 5 段式写法(Seedance / 小云雀 / Sora / 可灵 / 即梦通用),含 14 条翻车规避规则。挂给写视频提示词、分镜、短剧脚本的步骤。.
ao-workflow-runner
多角色 YAML 工作流执行引擎——解析 workflow YAML,加载 agency-agents-zh 角色,按 DAG 顺序执行.
subagent-driven-development
Execute plans via delegatetask subagents (2-stage review).
delegate-to-dsh
Route bounded coding work from Codex to DeepSeek Harness through a named model profile and an isolated worktree. Use for explicit or conversational requests such as “use DSH”, “let DSH handle this”, “交给 DSH”, “使用 DSH 处理”, alternate-model execution, or work that should return a reviewable patch and evidence.
setup-dsh-bridge
Set up, repair, or reconfigure DSH Codex Bridge for the current Git project. Use when the user installs the plugin, asks to configure DSH, add or switch a model, change the default Profile, preview model configuration, roll back a Profile change, or when delegation reports missing/invalid Bridge configuration.
Subagent Driven Development
Orchestrate specialized, autonomous AI agents to execute parallel subtasks with strict boundaries.