Sylars Work Manager — Agent Operations Workspace
Give agents fresh context and bounded workspaces—then keep consequential decisions human.
Context Boundary
CLI Providers
Approval Gate
Tech Stack
The Problem
Repository automation becomes unreliable when an agent starts from stale documentation, works against the wrong revision, or receives credentials broader than the task requires. The operational problem is not merely invoking a model; it is proving what context it saw, constraining where it can work, and preserving a safe review boundary before changes move forward.
The Challenge
Sylars needed to coordinate long-running agent work without turning the workspace into an opaque autonomous system. Every task therefore begins with origin and freshness checks, receives an isolated clone at the exact default-branch SHA, and records context attestations. Approval resume fails closed if the remote has advanced, while repository credentials remain outside the agent sandbox.
Architecture & System Design
An Azure VM runs the workspace UI, Hermes agent runtime, provider gateway, automation runner, project intelligence services, and Sylars control plane as pinned Docker Compose services behind Caddy. Signed GitHub push ingestion creates exact-SHA code and documentation generations. REST and MCP interfaces expose authenticated project context, while the control service manages threaded task submission, progress, steering, cancellation, and approval-gated continuation.
Results
The production stack is deployed with health-checked services, immutable upstream container digests, scheduled upgrade proposals, persistent provider authentication, exact-SHA freshness enforcement, and bounded automation for project digests, issue triage, PR and CI scans, and stack health. The system keeps raw provider output private while surfacing concise, actionable notifications and validated links.
More from Personal Project
Ummidvar — AI Job Application Agent
Autonomous, multi-profile job application agent: discovers roles across 7+ boards, ranks each 0–100 with a 5-component explainable scorer, generates hallucination-safe tailored applications via a Facts Graph, auto-submits through real ATS systems, tracks replies via Gmail, and flags visa-sponsorship signals across 5 countries — 286 tests, 11 modular packages, zero vendor lock-in.
Ummidvar — Autonomous Job Agent
Product vision and feature scoping for Ummidvar — an autonomous, multi-profile job application agent. Defined the 11-package modular architecture from a product perspective, prioritised the anti-hallucination Facts Graph as a non-negotiable product constraint, and scoped visa-sponsorship intelligence for 5 international markets.
Watchdog — Multi-cloud Observability Control Plane
A privacy-minimised observability and safe-operations control plane spanning AWS, Azure, GCP, container hosts, registry storage, incidents, and exact human-approved cleanup plans.
Swadhika Studios — Editorial Interior Portfolio
A calm, editorial website for an interior-design and turnkey-delivery studio, built around full-bleed project photography, deliberate typography, and unhurried motion.
BidWave — Freelance Opportunity Intelligence
A self-hosted opportunity engine that aggregates freelance listings, ranks fit in two stages, drafts tailored proposals, and keeps application decisions under human control.
RenderBoard AI — Interior Design Workspace
An AI-image-first interior design workspace for plan intake, review, layout, rendering, mood boards, and export—with safety, billing, and human approval designed into the flow.
Atelier CRM — Studio Readiness & Delivery
A status-first CRM for boutique interior-design studios, connecting client approvals, procurement readiness, installation milestones, and project delivery in one operating view.
Interested in this work?
Full architecture walkthrough and code review available during interviews.