hephaestus

hephaestus is a cursor rule for coding agents from agentlas-ai/Agentlas-OS. It costs 553 tokens per session, scanned A, original, Apache-2.0.

Hephaestus Workforce routing and mandatory multi-turn roster continuity through local Agentlas OS Core.

Cursor rule

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add rules/agentlas-ai/agentlas-os/hephaestus
Clone the repo
git clone --depth 1 https://github.com/agentlas-ai/Agentlas-OS

Wrote 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.

agentmods badge for hephaestus

README.md
[![agentmods](https://agentmods.dev/badge/rules/agentlas-ai/agentlas-os/hephaestus.svg)](https://agentmods.dev/rules/agentlas-ai/agentlas-os/hephaestus)
Your own site
<a href="https://agentmods.dev/rules/agentlas-ai/agentlas-os/hephaestus"><img src="https://agentmods.dev/badge/rules/agentlas-ai/agentlas-os/hephaestus.svg" alt="Measured on agentmods" height="20"></a>
Per session 553 This file is loaded in full into every session.
When invoked 553 The same file — it is already loaded in full.
Security scan A 0 findings. Scan, not verified.
Origin unknown No closer match found in the catalogue.
Token cost

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.

ModelPer sessionOnce invoked
Fable 5 $0.00553 $0.00553
Opus 5 $0.00277 $0.00277
Sonnet 5 $0.00111 $0.00111
Haiku 4.5 $0.00055 $0.00055

Measured today against content hash 3c04400b11b4, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

hephaestus scanned grade A with 0 findings 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 today.

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.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

cursor/plugin/rules/hephaestus.mdc · 47 lines

What it actually says

Hephaestus Network (Cursor fallback)

Cursor ships explicit /hep-network, /hep-local, /hep-cloud, and /hep-hub commands. For /hep-network <request> or @Hephaestus <request>, use MCP server hephaestus-network, the only host-visible Workforce MCP. Core owns its Cloud/Hub upstream calls.

At the start of every nontrivial project turn, call workforce.goal_context({projectDir}). If an active binding covers the ongoing work, reuse that exact roster plus local skills before recruiting. The user does not need to say goal.

  1. Author a redacted agentlas.workforce-work-order.v1; keep private files, memory, secrets, and raw project grounding local. Use title/task plus only genuinely constraining communities/roles/skills/knowledge for semantic fit. Carry genuinely required tool capabilities, authorities, runtimes, languages, or modalities separately for post-selection host inventory and permission-receipt validation; they never rank or exclude semantic candidates. Omit unconstrained fields (absent = empty).
  2. Call workforce.search_candidates with {workOrder, sourceScope: "network"} and keep federationResult.
  3. Author the host-LLM Selection from content/qualification evidence. Call workforce.validate_selection with {workOrder, selection} and keep federatedSelection.
  4. Call workforce.prepare_execution with {workOrder, selection, federatedSelection, projectDir, goalId?}. Core requires projectDir, derives a goal id when needed, and automatically binds every successful preparation before execution.
  5. Run planner/manager, workers, synthesis, and verifier as distinct calls with explicit handoffs. Claim execution only from complete receipts.

On later turns record reuse|local-only|recruit|standby|blocked through workforce.record_goal_turn. Release a roster only after explicit whole-goal completion/cancellation with workforce.complete_goal(explicitCompletion=true). A 24-hour lease affects only the next Hub charge and never ends the binding. Standby is durable availability, not a continuously running model.

Network is registered Local + signed-in owner Cloud + public Hub. Missing sources stay explicit. Do not call hephaestus_route, bypass Core with direct remote search, accept deterministic staffing, or silently substitute workers.

Changes

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.

  1. today First seen · 47 lines · 553 tokens per session scan A 3c04400b11b4

Subscribe to this mod's changes

hephaestus is a cursor rule published in the GitHub repository agentlas-ai/Agentlas-OS (1,099 stars, last pushed yesterday), licensed Apache-2.0. It adds 553 tokens to every session, about $0.0028 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-09-03.