Changelog

All notable changes to antidote are documented here. The format follows Keep a Changelog, and this project adheres to Semantic Versioning.

[1.1.0] - 2026-08-20

Added

• .cursor-plugin/plugin.json — native Cursor plugin install support.
• .devin-plugin/plugin.json — native Devin plugin install support.
• .grok-plugin/plugin.json and .grok-plugin/marketplace.json — native Grok Build plugin install support.
• plugin.json (repo root) — canonical plugin manifest for Antigravity, Qwen Code, Factory Droid, and other root-native agents.
• repository field added to all adapter manifests.
• metadata block (author, source, version) added to SKILL.md frontmatter per the Agent Skills spec.
• docs/benchmark.svg — visual benchmark chart with model logos and grouped bar comparison.
• Cursor, Devin, and Grok Build install sections in README.
• Cursor, Devin, and Grok Build entries in docs/agent-portability.md.

Fixed

• Claude Code skill discovery: marketplace source changed from ./skills/root-cause to ./ so Claude Code finds skills/root-cause/SKILL.md via the standard skills/ convention. Previously the plugin installed but exposed no skills.
• Removed fabricated "75+ agents" claim from README and docs.
• Removed false "search for root-cause in the Cursor plugin marketplace" claim (not published there).

[1.0.0] - 2026-08-18

Added

• skills/root-cause/SKILL.md — the skill prompt. Evaluated with the skill-eval-harness across five models (Claude Haiku 4.5, Claude Sonnet 5, GLM 5.2, tencent/hy3, moonshotai/kimi-k3). On a 4-case benchmark where a script oracle compiles and runs the model's fix to check it is structural (not a patch), the skill raised the structural-fix pass rate from 0% to 56% on Sonnet 5 and GLM 5.2, 0% to 50% on kimi-k3, 0% to 44% on hy3, and 31% to 50% on Haiku.
• Adapter manifests for Claude Code, Codex/ChatGPT, and GitHub Copilot CLI.
• docs/agent-portability.md — host adapter reference.
• docs/benchmark.md — evaluation methodology and per-model results.
• CI workflow validating the skill against the Agent Skills spec via skills-ref.
• MIT license.