Dev collapses Anthropic's 5-piece always-on agent stack into one Claude Code prompt
InternationalData569 · reddit · 2026-11-17
A developer packed all five pieces of infrastructure from Anthropic's 'Building always-on agents' guide—event inbox, timers, structured memory, permission guardrails, and an outbox—into Quarterdeck, an open-source Claude Code plugin backed by local SQLite with zero servers or YAML.
- Inbox & Timers: watches shell commands, git/GitHub/GitLab or MCP connectors on cron schedules, with single-worker serialization and crash recovery
- Memory + 'dreaming' curation: each wake-up loads a structured brief (objective, pinned rules, open items, run summaries); pinned rules stay locked against drift
- Deterministic guardrails: a pre-tool hook intercepts external writes and taints runs touching untrusted inputs (public issues, emails), routing risky actions to an approval queue (off/propose/auto modes)
- Outbox: quiet runs by default; pings your phone via Claude Code Remote Control only when a human decision is needed
A bundled /goal-new setup wires four coordinated goals—Issue Refiner → Issue Fixer → PR Reviewer → CI Fixer—using issue labels as a state handshake to avoid races.
More from coding & agent
- sandbox-mcp: Hanselman's MCP server gives AI agents disposable real Windows 11 machines — unixterminal · 2026-11-17
- ProceduralTerrains: open-source Three.js GPU terrain generator hits 2.0 — anselm · 2026-11-17
- Zed founder on agent skills: absurdly slow, quality all comes from babysitting — zeeg · 2026-11-17
- Dev open-sources inference-optimisation agent skill he says made him $100K — Kyrannio · 2026-11-17
- Open-source Hermes Agent profile: cheap main model plus low-cost side-task pinning — alexcovo_eth · 2026-11-17
- CUDA guide details fine-grained sync: sentinel-based tile readiness over membar, faster on Rubin — simran_s_arora · 2026-11-17