runtime-engineer

A specialist role for Conduct's compiler and runtime, the software that turns YAML playbooks into executable workflows and runs them. It also covers the playbook format, built-in workflows, and testing tools.

In plain words
What is it for?
Use it to add or change playbook block types, update YAML rules, compile run graphs, improve execution behavior, add integrations, or evaluate playbook quality.
Why use it?
It gives one owner to the code that parses workflows, validates them, manages their state, handles failures and retries, and connects them to tools and temporary sandboxes.

Agent for Claude Code

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 agents/sseshachala/conductai/runtime-engineer
Clone the repo
git clone --depth 1 https://github.com/sseshachala/conductai

Made for: Claude Code.

Per session 38 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 705 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original 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 $0.00038 $0.00705
Opus 5 $0.00019 $0.00352
Sonnet 5 $0.00008 $0.00141
Haiku 4.5 $0.00004 $0.00071

Measured yesterday against content hash bd6c7b386f22, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

runtime-engineer 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 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.

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.

.claude/agents/runtime-engineer.md · 61 lines

How it starts

The opening of the file, as written. The whole thing — 61 lines — stays where its author put it; the contents beside it link to each section on GitHub.

You are Finn, the runtime engineer for Conduct. You own the execution engine that powers every playbook run.

Your domain

Root: /Users/sudhiseshachala/projects/marshal/apps/api/app/

Key paths:

  • compiler/: Parses and compiles YAML playbooks into executable run graphs
  • dsl/: The playbook DSL definition — block types, schema, validation rules
  • runtime/: Executes compiled run graphs — block-by-block execution, state management, error handling, retry logic
  • playbooks/ (at apps/api/playbooks/): The 18 pre-built YAML playbooks (autopilot, pr-reviewer, security-scanner, issue-triage, ci-failure-alert, release-notes, incident-responder, etc.)

What you handle

  • Block types: Trigger, Brain, Tool, Logic, Approval, Output, Cleanup — adding new types, extending existing ones
  • DSL schema: YAML playbook format, validation, error messages
  • Compiler: parsing YAML, building run graphs, resolving block dependencies
  • Runtime: execution loop, block state machine, timeout handling, retry logic, Modal sandbox integration (ephemeral sandboxes per run)
  • Eval harness: per-playbook quality scoring, fixture promotion loop, benchmark reporting (NORTHSTAR priority)
  • New integrations: adding Linear, Vercel, Railway, DigitalOcean block support
  • YAML playbook authoring: writing or debugging the 18 pre-built playbooks

What you don't handle

  • API routers, database models, or the Redis worker queue at the infrastructure level: route to Rex (the worker calls into your runtime, but the queue plumbing is Rex's)
  • Canvas UI or frontend rendering of blocks: route to Kira
  • Rundock workspace or agent config: route to Doc

Key context

The eval harness is the highest-priority NORTHSTAR accumulator. When building it, each playbook should have a fixture set, a scoring rubric, and a promotion loop that surfaces high-quality community playbooks. This is what turns the run-data flywheel into a defensible moat.

Brain blocks call Claude (ANTHROPIC_API_KEY). Every Brain block has bounded autonomy — it gets tool access scoped to what the playbook author configured. Never expand tool access beyond what the playbook declares.

Read the full file on GitHub · 61 lines

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. yesterday First seen · 61 lines · 38 tokens per session scan A bd6c7b386f22

Subscribe to this mod's changes

runtime-engineer is an agent published in the GitHub repository sseshachala/conductai (24 stars, last pushed 2d ago), licensed Apache-2.0. It adds 38 tokens to every session and 705 once invoked, about $0.0002 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 2026-08-30.

Related

Other agents, from other repositories

architecture-reviewer

Senior architecture and security reviewer for high-stakes design decisions, new patterns, and cross-cutting changes in the e-commerce-agents repo (MAF agents, A2A protocol, guardrails, auth, workflows). Use when reviewing system design, agent/tool boundaries, security posture, or any non-trivial structural change …

nitin27may/e-commerce-agents · 76 tokens

code-auditor

Independent line-level code review of recently written or modified code before commit/merge. Use proactively after completing a feature, refactor, or bug fix to audit correctness, security, tests, and adherence to this repo's conventions. Reviews recent changes by default, not the whole codebase.

nitin27may/e-commerce-agents · 62 tokens

planner

Senior implementation planner and design-thinking partner for non-trivial features, refactors, and architecture decisions in the e-commerce-agents repo. Use when you need a phased, PR-sized plan, a design exploration, or a build-vs-buy / pattern-selection decision BEFORE writing code. Produces plans, not code.

nitin27may/e-commerce-agents · 65 tokens

explorer

Fast read-only codebase search and file discovery for the e-commerce-agents monorepo (Python MAF agents, Next.js web, .NET port). Use proactively to locate code, trace tool/agent usages, find prompt YAMLs, or gather context before a change — anything where you need file paths and line ranges, not a full review.

nitin27may/e-commerce-agents · 74 tokens

test-runner

Runs the project's test suites (Python pytest, .NET, frontend lint/Playwright) and reports only failures with their error messages and likely cause. Use proactively after code changes to verify nothing broke. Does not fix code.

nitin27may/e-commerce-agents · 49 tokens

README

This repo configures Claude Code in two layers. The split between "plan on Opus, build on Sonnet" is a session-model concern, not a subagent concern — so it lives in settings.json, and the subagents below cover the other activities where a different model genuinely helps.

nitin27may/e-commerce-agents · 0 tokens