Skillscript: Empowering Agents for Local Repetitive Tasks
sshwarts · reddit · 2026-07-12
The author introduces Skillscript, a small declarative language for writing automated workflows for AI agents. The goal is to offload "trivial, well-defined" tasks to local models and deterministic steps rather than repeatedly burning frontier model tokens.
The core idea includes:
- Once a task is written as a skillscript, many steps become deterministic operations (like fetching, filtering, counting, branching) that consume almost zero model calls.
- Only parts that genuinely require understanding and decision-making are handed over to local models or minimal model inference.
- Frontier models are only used once to compile the task into a skillscript; subsequent periodic executions remain low-cost and locally runnable.
It also emphasizes trust boundaries: skills are initially generated as drafts and can only run after human approval; an agent cannot approve itself, and this limitation is a cryptographic boundary, not just a standard policy setting. The author notes the project is in its early stages but already usable in daily work, and they are curious how others decide which steps to keep local.
More from coding & agent
- AI agents are starting to strain code hosting platforms — craigsdennis · 2026-07-21
- Async OPD distillation doubles throughput while matching synchronous math accuracy — _lewtun · 2026-07-21
- Omnigent 0.6.0 adds Claude Code imports, Slack approvals and desktop apps — matei_zaharia · 2026-07-21
- Google appears to have quietly shipped Gemini 3.6 Flash, with lower pricing and better agentic scores — xiaohu · 2026-07-21
- Open-source CLI audits AI tools, MCP configs, and agent skills on local machines — Initial-Copy332 · 2026-07-21
- Coding agents feel less stressful when the 5-hour limits are temporarily removed — iamrobotbear · 2026-07-21