agent-hub
Skill1 starsChanged 9 months agoMITView on GitHub
Plug-and-play orchestration skill. User states a goal in natural language; the model runs commands/agent-hub.md (dispatcher), classifies work via agent/operating-model.md, then opens only the matching command or architecture doc (map, architect, memory, recovery, verify, next-pass, or one NN-layer). Layered docs for intent, policy, context, planning, routing, delegation, execution, MCP/tool boundaries, verification, memory, recovery, observability, human gates, release. Use for repo orchestration reviews, boundary work, memory/recovery design, permission-sensitive refactors.
2 other projects use a version of this.
Adopted by
Most recently taken up by
- gmannerHJ/apify-open-data-actorstheir own version, first seen Nov 1, 2026
- LakshmiSravyaVedantham/agent-hubtheir own version, first seen Oct 9, 2026
Other versions2 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 2
- LakshmiSravyaVedantham/agent-hub★ 1same name
- gmannerHJ/apify-open-data-actors★ 0same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository15 other files it gives its agents.
- .cursor/rules/AGENT.mdcCursor rule
- .cursor/rules/AI Text Humanizer Specialist.mdcCursor rule
- .cursor/rules/App_reviewer.mdcCursor rule
- .cursor/rules/maintainer.mdcCursor rule
- .cursor/rules/production.mdcCursor rule
- .cursor/rules/Project_Manager.mdcCursor rule
- .cursor/rules/README.mdcCursor rule
- .cursor/rules/rules.mdcCursor rule
- blueprintkitSkill
- context-engineeringSkill
- core-engineeringSkill
- co-star-uiSkill
- erc8004-agent-creatorSkill
- startupSkill
- webmcp-agentsSkill
The file
We split this skill into the rules it gives an agent
5 rules. About 0 are common, 1 is distinctive, and 4 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
This author's own choices. Worth a look.
4
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 13
Plug-and-play: Start from commands/agent-hub.md when the user’s goal is broad or ambiguous.
- Distinctive: line 27
Do not assume a particular framework, vendor SDK, or single entry file name in application code.
- Very specific: line 28
Do not assume one universal verification command covers all quality gates.
- Very specific: line 29
Do not assume memory is one subsystem.
- Very specific: line 30
Do not assume UI control flow equals MCP transport, client logic, or tool execution.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
agent-hub (machine-readable entrypoint)
What this is
agent-hub is a documentation-driven control plane for agents: layered architecture, indexes, diagrams, templates, command specs, and explicit editing rules. Application code under review may live outside this folder; this package defines how to navigate and change it safely.
Plug-and-play: Start from commands/agent-hub.md when the user’s goal is broad or ambiguous. Infer intent, then route to the smallest command or single architecture/NN-*.md.
Canonical path: SKILL.md at the root of this package (this file).
Read first (minimum)
agent/README.md— operating scope for agents using this package.agent/operating-model.md— classify tasks and pick a route.architecture/00-scope-and-rules.md— global rules and doc authority.
Then branch by task using indexes/orchestration-map.md and the matching architecture/NN-*.md file. For delegated specialist work, see subagents/README.md and run routing via commands/agent-hub.md.
What not to assume
- Do not assume a particular framework, vendor SDK, or single entry file name in application code.
- Do not assume one universal verification command covers all quality gates; verification is multi-entry (skill, command, workflow). See
architecture/07-verification-and-reflection.md. - Do not assume memory is one subsystem; session, durable extraction, and shared team memory differ in triggers and writers. See
architecture/08-memory-pipelines.md. - Do not assume UI control flow equals MCP transport, client logic, or tool execution. See
architecture/06-mcp-and-tool-boundaries.md.
Where to look
| Concern | Primary doc |
| Orchestration overview | indexes/orchestration-map.md |
| Intent and commands | architecture/01-intent-and-commands.md |
| Policy and permissions | architecture/02-policy-and-permissions.md |
| Context and state | architecture/03-context-and-state.md |
| Planning and routing | architecture/04-planning-and-routing.md |
| Delegation and execution | architecture/05-delegation-and-execution.md |
| MCP and tool boundaries | architecture/06-mcp-and-tool-boundaries.md |
| Verification and reflection | architecture/07-verification-and-reflection.md |
| Memory pipelines | architecture/08-memory-pipelines.md |
| Recovery and resume | architecture/09-recovery-and-resume.md |
| Observability and human gates | architecture/10-observability-and-human-gates.md |
| Release and scope | architecture/00-scope-and-rules.md |
| Specialist subagents | subagents/README.md |
| Cursor launcher setup (paste into chat) | COMMANDS.md |
High-risk areas (typical patterns)
- Central mutable session or bootstrap state: guarded boundary; do not add fields casually. See
architecture/03-context-and-state.md. - Main query or agent loop modules: high blast radius; prefer phased extraction. See
architecture/05-delegation-and-execution.md. - Permission and human-gate hooks: easy to double-resolve or bypass; coordinate with
architecture/02-policy-and-permissions.mdand10-observability-and-human-gates.md. - MCP clients and connection managers: easy to conflate with UI; see
06-mcp-and-tool-boundaries.md.
Choosing commands vs docs vs agent guidance vs subagents
- Universal dispatcher:
commands/agent-hub.md(routing modes, when to assign a specialist first). - Quick launcher: other
commands/*.md, or optional workspace.cursor/commands/agent-hub-*.md/agent-hub.mdwhen present (seeindexes/command-index.md). To create those files automatically, the user can attachCOMMANDS.mdand follow its paste prompt. - Specialist persona:
subagents/*.mdfor scoped analysis or planning; they link into canon instead of replacing it. - Deep boundary rules:
architecture/*.md. - Behavioral contract (stop conditions, escalation):
agent/*.md. - Full audit:
commands/architect.mdworkflow, thentemplates/inspection-pass-template.md.
Commands index
See indexes/command-index.md for canonical commands/*.md versus optional editor launchers.
Decisions
indexes/decisions-index.md— architecture decision log pointers.docs/decisions/README.md— how to record decisions.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.