Skip to content

Changelog

Jump to bottom
github-actions[bot] edited this page Jun 7, 2026 · 83 revisions

Latest

Changelog

This page summarizes recent wiki-relevant releases. The authoritative release history remains the repository CHANGELOG.md, GitHub Releases, and git tags.

Note

Wiki summaries intentionally focus on user-facing documentation changes and workflow boundaries. Use the repository changelog for exact release notes.

v2.21.7 · Issue Tracking and Tracer-Bullet Planning

Adds issue-comment discipline, context-aware requirement grilling, vertical-slice breakdown wording, and a TDD tracer-bullet guide.

Visible documentation points:

  • guides/templates/issue-tracking-comments.md provides pickup, progress/blocker, and completion comment drafts for issue-based work.
  • /requirements now challenges glossary conflicts, fuzzy terms, scenarios, code contradictions, and ADR conflicts when repo context exists.
  • /breakdown now prefers backlog-bound tasks that are independently verifiable vertical slices.
  • /build-brief can load guides/tdd-tracer-bullet.md for TDD, red-green-refactor, or test-first work.
  • Codex update-flow docs and GitHub Actions/link-check maintenance commits since v2.21.6 are included in this tag.

Boundary: markdown guidance only. No automatic issue mutation, setup engine, test runner, runtime enforcement, or dynamic orchestration.

v2.21.6 · Codex SessionStart JSON Compatibility

Fixes Codex v0.137.0 startup compatibility for the package SessionStart hook.

Visible documentation points:

  • hooks/session-start.sh returns valid JSON with hookSpecificOutput.additionalContext when Codex invokes the hook.
  • Claude/default runs still emit the markdown reminder directly.
  • tests/test-verbosity-hook.sh now includes a Codex JSON parse smoke test.
  • Compatibility docs now describe this as a narrow output adapter rather than Claude hook feature parity.

Boundary: hook-output compatibility only. No runtime enforcement, policy authorization, dynamic orchestration, or general Claude hook port to Codex.

v2.21.5 · Project Context Contract

Adds optional repo-local context guidance so skills can read glossary, issue-tracker, triage-label, domain-doc, and ADR context before reasoning.

Visible documentation points:

  • New guides/project-context-contract.md defines optional context files such as CONTEXT.md, CONTEXT-MAP.md, and docs/agents/*.md.
  • /requirements, /design, /build-brief, /diagnose, and /scrutinize now check glossary/context files when present.
  • /breakdown and the AGENT-BRIEF template now classify backlog-bound work as ready-for-agent, ready-for-human, or needs-info.
  • llms.txt indexes the guide for cross-agent discovery.

Boundary: markdown guidance only. No setup engine, automatic issue mutation, runtime enforcement, marketplace behavior change, or Claude hook port to Codex.

v2.21.4 · Release Decision Gate

Adds explicit release classification so maintainers choose release now, bundle later, or no release before version files change.

Visible documentation points:

  • /deploy-guide now includes a plugin release decision gate.
  • CONTRIBUTING.md mirrors the same three categories in the release checklist.
  • tests/validate-content.sh pins the gate so the workflow does not drift back into implicit release decisions.

Boundary: markdown workflow guidance and validation only. No runtime enforcement, marketplace behavior change, skill automation, or Claude hook port to Codex.

v2.21.3 · Codex Project Context Files

Adds concise repo-local context for Codex and other non-Claude agents.

Visible documentation points:

  • DOMAIN.md captures invariants, safety rules, data boundaries, and validation expectations.
  • SPEC.md provides a fast project digest and session re-entry context.
  • .codex/README.md documents repo-local Codex setup and the Claude-hook boundary.
  • docs/adr/README.md and ADR-025 record the ADR index and project-context decision.
  • AGENTS.md and llms.txt point agents at these context files.

Boundary: documentation and session ergonomics only. No runtime enforcement, marketplace behavior change, skill behavior change, or Claude hook port to Codex.

v2.21.2 · Codex Installed-Cache Validator Context

Fixes the Codex installed-cache validator context without changing skill behavior.

Visible documentation points:

  • tests/validate-structure.sh now accepts the installed-cache shape where Codex omits plugin -> . because the installed root is already the plugin root.
  • Source and marketplace snapshots still require plugin -> . for publishability, because the Codex marketplace descriptor points at ./plugin.
  • docs/codex-integration.md explains when to run source/marketplace validation versus installed-cache validation.

Boundary: packaging validation and documentation only. No runtime enforcement, marketplace behavior change, skill behavior change, or Claude hook port to Codex.

v2.21.1 · Cross-Agent Evidence Discipline

Adds explicit limitations and release-evidence guidance for cross-agent users.

Visible documentation points:

  • New Limitations page explaining where the plugin helps, what it does not enforce, and what evidence belongs in PR/release proof.
  • docs/compatibility-matrix.md now compares shared markdown skills, Claude Code packaging, Codex packaging, hooks, memory, enforcement, and release evidence.
  • CONTRIBUTING.md now asks PRs that affect user-facing doctrine, install, release, generated catalog, or runtime-boundary surfaces to include real behavior proof.
  • tests/validate-structure.sh now keeps the Limitations page in the required wiki skeleton.

Boundary: documentation and release discipline only. No runtime enforcement, package behavior change, marketplace behavior change, or Claude hook port to Codex.

v2.21.0 · Cross-Agent Discovery and Portability Contract

Adds a generated skill catalog and documents the shared SKILL.md frontmatter contract for Claude Code, Codex, and other markdown-capable agents.

Visible documentation points:

  • docs/data/skills.json is generated from skills/*/SKILL.md for cross-agent discovery.
  • CONTRIBUTING.md, guides/skill-authoring.md, and docs/compatibility-matrix.md document required, optional, cross-agent, and Codex-ingestible frontmatter fields.
  • docs/codex-integration.md and llms.txt point tools to the generated catalog.
  • guides/structured-output-protocol.md adds a compact handoff-integrity note pattern.
  • /review-ai, /reflect, and guides/quick-reference.md add observable AI-work health signals such as loops, retries, context compaction, audit evidence, and next-session recovery.

Boundary: generated metadata and markdown guidance only. No runtime dispatcher, no Claude hook port to Codex, no budget enforcement, no policy gate, and no agent-to-agent orchestration protocol.

v2.20.2 · Production Canary Reconciliation Gates

/deploy-guide now covers provider-managed production canaries and capacity changes where the requested canary target and the provider-selected target may differ.

Visible documentation points:

  • Precheck, cordon, observation, drain, provider-side change, reconciliation, and postcheck phases.
  • Planned target vs actual provider-selected target comparison.
  • Desired/min/max capacity, readiness, schedulable capacity, and unintended SchedulingDisabled checks.
  • Unresolved rollout state routes to /operational-state.

Boundary: no cloud execution, policy enforcement, Kubernetes automation, ASG automation, or runtime state engine.

v2.20.1 · Incident/Config Consistency-Lite

/consistency-check now includes a lightweight mode for incident and config hotfix work that does not have persisted spec artifacts.

Visible documentation points:

  • Checks symptom, evidence, root cause, actual fix, deploy path, live verification, and drift.
  • Flags overclaiming PR/changelog text, missing evidence, scope mismatch, deploy drift, and unclassified adjacent operational state.
  • Routes unresolved related findings to /operational-state.

Boundary: no runtime enforcement, cloud execution, alert mutation, or automatic issue closure.

v2.20.0 · Operational State Model

Adds /operational-state, a read-only classifier for operational findings.

States:

  • Watch
  • Fix Candidate
  • Active Incident
  • Resolved
  • Handoff
  • Known Accepted Issue
  • False Positive
  • Self-Resolved

The skill maps each state to evidence, allowed and prohibited actions, approval gates, artifacts, escalation criteria, and closure criteria.

Boundary: no runtime state engine, policy enforcement, cloud execution, alert suppression automation, or automatic production write.

v2.19.2 · Operational Doctrine Patch

Ships doctrine-only improvements to existing skills:

  • /deploy-guide classifies deploy type before rollout planning.
  • /security-check covers more infrastructure and configuration surfaces.
  • /reflect captures more granular skill-effectiveness feedback.
  • /management-talk includes an operational incident closure example.

Broader operational model work was intentionally deferred until later releases.

v2.19.1 · Codex Runtime Compatibility Contract

Clarifies that Codex can install the plugin and load the same markdown skills, but does not run Claude hooks or gain runtime enforcement.

Related docs:

v2.19.0 · Native Codex Plugin Packaging

Adds Codex packaging through .codex-plugin/plugin.json and .agents/plugins/marketplace.json while preserving the Claude Code package.

Install path:

codex plugin marketplace add pitimon/8-habit-ai-dev
codex plugin add 8-habit-ai-dev@pitimon-8-habit-ai-dev

Boundary: same read-only markdown skills, no Claude hook parity, and no runtime enforcement.

Earlier Releases

For earlier versions, use:

See Also

Source of truth: this wiki is generated from docs/wiki/. Edits made through the GitHub Wiki web UI may be overwritten by the next sync. To change a page, open a PR against the repository source file.

Repository · Issues · README · License

8-Habit AI Dev

Workflow discipline for AI-assisted development

Start

Workflow

Operations

Reference

Project

Clone this wiki locally