Keeping Specs in Repo Isn't Always Useful for Agents
zeeg · x · 2026-07-14
The author shares findings from their experiment of keeping specs long-term in the repo for co-location: it didn't provide enough value.
The main issues are:
- High maintenance overhead, with much content devolving into documentation rather than executable constraints
- Frequent updates lagging, causing agents to misjudge the "source of truth"
- Even when intent is documented, models often ignore it, leading to more manual copy-pasting
They plan to transition to a lighter structure:
- policies/: Only for very lightweight rules, like where code comments should go
- openspec/: Kept temporarily, but no longer used as context docs
- custom skills: To supplement third-party docs, like Vercel Sandbox and Slack API
- agents.md: Serves primarily as a resource map
- Using inline code comments to express intent
Related event: Experiment Shows Limited Value of Repo Specs for Agent Constraints(2 posts)→
More from coding & agent
- Kimi K3 rises to No. 4 on the Agent Arena leaderboard — HeyZoyaKhan · 2026-07-22
- Claude adds screen-recorded skills that can replay your workflow — CodeByPoonam · 2026-07-22
- Devin adds e2b sandboxes for remote agent execution — badphilosopher · 2026-07-22
- Hermes Agent Refactoring Proposal: Decoupling via Event Bus and Monorepo Slicing — Promptmethus · 2026-07-22
- ty now reads Pydantic config keywords and field metadata — charliermarsh · 2026-07-22
- Pensar Launches AI Security Agent to Autonomously Discover and Patch 0-Days — andriy_mulyar · 2026-07-22