OPENSPEC PUB_DATE: 2026.04.12

OPENSPEC V1.3.0 LANDS; RUNTIME MCP TRUST SCORES DEBUT

OpenSpec 1.3.0 broadens IDE/assistant support and fixes rough edges, while Dominion Observatory introduces runtime trust scores for MCP servers. [Fission-AI/Op...

OpenSpec v1.3.0 lands; runtime MCP trust scores debut

OpenSpec 1.3.0 broadens IDE/assistant support and fixes rough edges, while Dominion Observatory introduces runtime trust scores for MCP servers.

Fission-AI/OpenSpec v1.3.0 adds tool/command generation for JetBrains Junie, config support for Lingma IDE, and tool support for ForgeCode and IBM Bob. It also makes shell completion install opt-in to avoid PowerShell encoding issues, stops false Copilot detection from a bare .github/ folder, fixes pi.dev command generation, updates the OpenCode adapter to .opencode/commands/, and makes openspec status exit cleanly when nothing changed.

Separately, Why Static MCP Quality Scores Are Not Enough introduces the Dominion Observatory, which tracks 4,400+ MCP servers with behavioral data. It exposes an MCP endpoint and REST APIs (trust checks, reporting, leaderboard, compliance) so agents can choose servers based on live performance, not static badges.

[ WHY_IT_MATTERS ]
01.

Agent toolchains get wider IDE/assistant coverage and fewer integration footguns.

02.

Runtime trust scoring gives teams a concrete signal to pick reliable MCP servers under real load.

[ WHAT_TO_TEST ]
  • terminal

    Upgrade to OpenSpec 1.3.0 and validate shell completions across Bash/Zsh/PowerShell; confirm no PowerShell encoding issues and correct Copilot auto-detection on repos with only .github/.

  • terminal

    Call Dominion’s trust API for your MCP servers, then POST reports from production agents for a week to see if trust scores track observed latency/error rates.

[ BROWNFIELD_PERSPECTIVE ]

Legacy codebase integration strategies...

  • 01.

    If you use OpenCode, update pipelines to the .opencode/commands/ directory and adjust any tooling that assumed the singular path.

  • 02.

    Introduce a trust gate before MCP calls: query Dominion first and fail over to higher-trust servers when scores dip.

[ GREENFIELD_PERSPECTIVE ]

Fresh architecture paradigms...

  • 01.

    Adopt OpenSpec early so one tool spec can target multiple IDE assistants without bespoke adapters.

  • 02.

    Design your agent client to report outcomes to Dominion and use trust scores in routing decisions from day one.

SUBSCRIBE_FEED
Get the digest delivered. No spam.