Skip to content

Releases: jqbit/yorient

v1.4.0 — post-release audit pass

25 May 18:52
@jqbit jqbit
v1.4.0
6ae6f6b

Choose a tag to compare

View all tags
v1.4.0 — post-release audit pass Latest
Latest

Post-release audit pass. Three parallel review tracks (content correctness, repo hygiene, ecosystem alignment) produced 30+ findings; this release folds the top 8 high-impact / low-effort items in. Backward-compatible with v1.3.0 — v=1 blocks unchanged; legacy v=0 blocks still upgrade in place.

Highlights

  • Algorithm hardened. ## Managed Block Algorithm now spells out the byte-level rules (BOM preservation, CRLF vs LF detection, trailing-newline preservation, malformed-input abort) and folds the placement rule into the zero-blocks-found branch.
  • Detection regex hardened. Portable across GNU grep, BSD grep, and ripgrep; tolerant of whitespace variants, CRLF, and the v=N attribute.
  • Ecosystem accuracy pass. Cursor wording softened (CLI-vs-IDE split was unverified); CLAUDE.md clarified as community convention; Gemini CLI/Antigravity global-config sharing noted.
  • agents.md citation. Names the standard, links the spec, frames yorient as layering atop it.
  • Community scaffolding. CONTRIBUTING.md, CODE_OF_CONDUCT.md (Contributor Covenant 2.1), SECURITY.md, issue + PR templates.
  • Dogfood. This repo now demonstrates yorient on itself — root YORIENT.md, AGENTS.md, CLAUDE.md, and a managed README block.
  • CI safety net. markdownlint-cli2 + lychee link checks on every push and PR (internal hard-fail, external warn-only).
  • Hygiene. .gitattributes, branch protection (linear history required), README badge row, sharper opening.

Full changelog

See CHANGELOG.md.

Assets 2
Loading

v1.3.0 — initial public release

25 May 18:25
@jqbit jqbit
v1.3.0
80a27a5

Choose a tag to compare

View all tags
v1.3.0 — initial public release

First public release of yorient — a Markdown-only orientation layer for AI coding agents.

yorient drops a canonical YORIENT.md at the repo/vault root and adds thin managed pointer blocks to README.md, AGENTS.md, CLAUDE.md, and GEMINI.md so every agent (Claude Code, Codex, Cursor, Hermes, Gemini/Antigravity, OpenCode, Pi, Factory/Droid, GitHub Copilot) reads from the same source of truth.

Highlights

  • Explicit managed-block algorithm. Idempotent insert/replace/de-duplicate rules. Re-running the skill on an unchanged file is byte-identical.
  • Versioned block markers (v=1). Future schema upgrades won't trample user content. Legacy unversioned blocks are detected and upgraded in place.
  • Worked example. A fully filled-in YORIENT.md for a hypothetical Node monorepo as a calibration target.
  • Ecosystem cheatsheet. Manifest files, layout conventions, and "read first" hints for Node/TS, Python, Rust, Go, polyglot monorepos, docs sites, and Obsidian vaults.
  • Drift check. Five-step verification recipe to keep the map honest as the repo evolves.
  • Smoke test. Per-agent installation verification.
  • Measurable subagent fan-out. File-count buckets tied to git ls-files | wc -l, with explicit token-budget and concurrency-limit guidance.

Install

Clone and copy the folder into each agent's skill tree — see README.md for the per-agent paths.

Migration from orient-map / ORIENT.md

references/rebrand-and-pointer-blocks.md includes detection commands, a three-branch decision tree, and an explicit string replacement map.

Full changelog

See CHANGELOG.md.

Loading