Skip to content
cccskills
FreePublished on GitHub

hep-hub

Staff a task only from public Agentlas Hub agents.

The English summary is being prepared. Showing the original description.

See how to install

Files included (1)

  • SKILL.md4.3 KB

SKILL.md (original)

Review the instructions the agent will receive before you install.

Update fallback: 자동 업데이트가 안 되면 hephaestus update를 한 번 실행하세요. 업데이트하지 않아도 현재 버전 명령은 그대로 동작합니다.

/hep-hub

Use the local Agentlas OS MCP server hephaestus-network and call the Workforce tools with exact sourceScope: "hub". This is public Hub only; it must not add registered Local or owner Cloud candidates.

First decide which of the two shapes this is

One named agent. The argument names a single agent — a slug, or an unmistakable name with no task described around it (ktx-book, use ktx-book). The user is asking to use that agent, not to have a team staffed, so do not make them sit through a staffing ceremony to get there:

  1. Find that exact agent with hephaestus_search. If nothing matches the name, say so and stop — never substitute a different agent for the one they asked for.
  2. Before preparing anything, tell them in plain words:
    • what the agent does, from its own card;
    • what it will need from them. Read prerequisites, envRequirements and mcpServers off the card. If all three are empty, say that the card declares nothing rather than implying it needs nothing — plenty of agents want a login they never wrote down, and the user should hear that before they start, not when the run stops to ask;
  3. Ask for whatever it needs, then prepare and run it, keeping the exact release pins below.

Skip the ceremony, never the pins: the prepared release must still carry source hub, the exact release, package hash, content digest, runtime bundle, permission policy, and context digest.

A job to be done. The argument describes work rather than naming an agent. Staff it:

  1. Author a redacted agentlas.workforce-work-order.v1; private project grounding stays on-host.
  2. Call workforce.search_candidates with {workOrder, sourceScope: "hub"} and keep the complete response as federationResult, including the Hub source receipt and the projected menu's selectionSessionId. Do not echo the projected menu back as federationResult — Core resolves the complete federation state locally from that session.
  3. Author the final agentlas.workforce-selection.v1 as the active host LLM, then call workforce.validate_selection with {workOrder, selection} and keep the response as federatedSelection. Revise on rejection; do not accept a deterministic picker or unrelated fallback.
  4. Call workforce.prepare_execution with {workOrder, selection, federatedSelection, projectDir}. Require every selected row to remain pinned to source hub, exact release, package hash, content digest, runtime bundle, permission policy, and context digest.
  5. Execute distinct planner/manager, worker, synthesis, and verifier calls with explicit artifact handoffs. Preserve packaged Team graphs.

Recurring Hub use

Public Hub packages are free to discover and call. Recurring work still uses the host's model, API keys, permissions, and scheduled execution capacity. Explain those prerequisites and obtain the user's scheduling instruction. Do not quote or purchase a retired Hub lease. The exact release remains bound to the goal until explicit completion or cancellation.

If the Hub source is unavailable or refuses the call, report its exact refusal; do not silently search Local or Cloud. Core owns the Hub upstream transport; do not expose a direct remote agentlas MCP alongside it. A prepared roster is not proof of execution.

If workforce.prepare_execution refuses with a repair object, the agent is broken, not absent. Say in plain words what repair.meaning says, offer the first available action in repair.repair, and run its command only after the user agrees. When a repair reports reselectionRequired: true, select again (workforce.validate_selection, then workforce.prepare_execution); never swap in a different agent without a new selection. A refusal that carries alternates lists other callable agents from the same search: present them as candidates and let the user or the selection step choose.

For a partial or failed result, report each source receipt's exact failureCode. Never collapse several receipts into one, substitute a different code, or relabel the outcome.

Raw request: $ARGUMENTS

Reviews

No reviews yet. Tell others what you thought.

Comments

No login needed. Share how it worked for you, ask a question, or note where it did not run.

No comments yet. Be the first to write one.

Comments are published right away. Do not include personal information such as email addresses or phone numbers. Up to 2 URLs.

0 / 1,000 characters

This form is protected by reCAPTCHA, and the Google Privacy Policy and Terms of Service apply.

More skills from this repository

Use when designing a new multi-agent team, visible agents folder, role boundaries, handoff flow, PM Soul, Memory Curator, Policy Gate, or evaluation role. Use for agent-team repo creation even when the user only says they want a meta-agent or agent operating system.

The English summary is being prepared. Showing the original description.

agentlas-ai/Agentlas-OS1.6KUpdated Oct 20, 2026

Single entry point for every Agentlas command: /agentlas network, /agentlas build, /agentlas storm, /agentlas graph, /agentlas one, /agentlas upload, /agentlas cloud, /agentlas hub, /agentlas local, /agentlas browser, /agentlas call, /agentlas search, /agentlas connect, /agentlas login, /agentlas orch, /agentlas update.

The English summary is being prepared. Showing the original description.

agentlas-ai/Agentlas-OS1.6KUpdated Oct 20, 2026

Single entry point for every Agentlas command: /agentlas network, /agentlas build, /agentlas storm, /agentlas graph, /agentlas one, /agentlas upload, /agentlas cloud, /agentlas hub, /agentlas local, /agentlas browser, /agentlas call, /agentlas search, /agentlas connect.

The English summary is being prepared. Showing the original description.

agentlas-ai/Agentlas-OS1.6KUpdated Oct 20, 2026

Use when adding or auditing local runtime behavior that turns a project folder into an Agentlas-aware workspace with .agentlas memory and sitemap files.

The English summary is being prepared. Showing the original description.

agentlas-ai/Agentlas-OS1.6KUpdated Oct 20, 2026

Use when the user types /agentlas-browser <url-or-query>, /agentlas browser, or /hep-browser for browser-required tasks, rendering pages, form flows, or visual evidence.

The English summary is being prepared. Showing the original description.

agentlas-ai/Agentlas-OS1.6KUpdated Oct 20, 2026

Use the Agentlas browser hardpoint for browser-required work.

The English summary is being prepared. Showing the original description.

agentlas-ai/Agentlas-OS1.6KUpdated Oct 20, 2026

See all skills by agentlas-ai

Report a problem with this skill