This release consolidates developer utilities into a unified plugin. The docs, prose, and ship plugins have merged into dev, providing changelog management, code review, and documentation workflows in one place.
The new /dev:review command spawns seven specialized reviewers in parallel to
analyze code changes with confidence-scored findings. Release workflows moved to
/dev:release and /dev:finalize. Documentation synchronization now works
across all doc commands with automatic worktree isolation.
💥 Breaking Changes
Section titled “💥 Breaking Changes”Consolidated docs and prose into dev plugin
Section titled “Consolidated docs and prose into dev plugin”Jan 22, 2026 · @mavam, @claude
The docs plugin has been renamed to dev and now includes the technical-writing skill from the former prose plugin. This consolidation creates a unified plugin for developer utilities.
Components:
- Skills:
dev:docs-authoring,dev:technical-writing - Agent:
@dev:docs-updater
Commands have been removed in favor of the @dev:docs-updater subagent.
🚀 Features
Section titled “🚀 Features”Unified developer utilities with integrated changelog and code review
Section titled “Unified developer utilities with integrated changelog and code review”Jan 22, 2026 · @mavam, @claude
The ship plugin has been consolidated into the dev plugin, creating a unified developer utilities plugin. The dev plugin now provides changelog management, code review, and release workflows alongside its existing documentation capabilities.
The @dev:changelog-adder agent replaces ship:adder for creating changelog entries. Seven specialized code reviewers (arch, docs, performance, readability, security, tests, ux) are now available under @dev:reviewers:*. The /dev:review command spawns these reviewers in parallel to analyze code changes with confidence-scored findings.
Release workflows moved to /dev:release and /dev:finalize commands. The changelog management features use tenzir-ship under the hood and support both standalone projects and module-based projects.
All external references in GitHub Actions workflows, the Tenzir plugin skills, and documentation have been updated to reference the new dev plugin agents and commands.
🔧 Changes
Section titled “🔧 Changes”Improved skill and agent descriptions with Use when pattern
Section titled “Improved skill and agent descriptions with Use when pattern”Jan 22, 2026 · @mavam, @claude
Improved YAML frontmatter descriptions across agents, skills, and commands in the dev plugin to follow the “Use when” pattern for better automated model calling and discovery.
The descriptions now explicitly state the use cases and trigger conditions (file types, commands, actions, topics) that help Claude determine when to invoke each skill or agent. This structured format enhances automated discoverability while maintaining clear, user-focused language.
Examples of improvements:
- Addressing PR comments skill: Changed from “Address GitHub PR review comments with commits” to “Address review feedback in comments of a GitHub pull request (PR)” with expanded trigger conditions
- All agents and skills: Descriptions now consistently follow the “Use when” pattern with specific trigger examples like checking error messages, reviewing complexity, validating input, or writing guides
This change applies the patterns defined in the project guidelines to ensure consistent and machine-parseable discovery across the entire dev plugin ecosystem.
Unified documentation synchronization across all doc commands
Section titled “Unified documentation synchronization across all doc commands”Jan 20, 2026 · @mavam, @claude
The documentation synchronization hook now runs across all three documentation commands (/docs:write, /docs:review, and /docs:pr) with an improved two-phase initialization approach. The first file operation automatically clones the docs repository if missing, while subsequent operations perform staleness checks. This change ensures consistent .docs/ synchronization across all documentation workflows without requiring manual git commands.
Additionally, command documentation now uses declarative language instead of explicit bash code blocks, making the instructions clearer and more maintainable.
Intelligent synthesis and structured tracking for review findings
Section titled “Intelligent synthesis and structured tracking for review findings”Jan 20, 2026 · @mavam, @claude
The review workflow now synthesizes findings using the orchestrator’s full context—all reviewer outputs, project information, and user intent—to distill actionable items. This replaces raw display with intelligent selection: deduplicating overlapping findings, correlating related issues, and filtering noise.
Reviewers assign structured issue IDs (SEC-1, ARC-1, TST-1, etc.) for tracking, and must provide concrete reasoning and evidence with each finding. The bar is raised at the source rather than adding post-hoc verification layers.
The orchestrator presents a focused summary sorted by urgency, then offers to enter plan mode where it re-reads full reviewer outputs for implementation planning.
Simplified severity display in review output
Section titled “Simplified severity display in review output”Jan 19, 2026 · @mavam, @claude
The review command now uses a simplified severity display that maps priority levels directly to colored emojis (P1→🔴, P2→🟠, P3→🟡, P4→⚪) instead of computing composite “action” labels from severity and confidence scores.
This change makes review findings easier to scan by eliminating the intermediate “critical/important/minor/optional” classification. Findings are now sorted by severity and confidence only, and the legend shows the direct P1-P4 mapping rather than derived action categories.
Structured hook outputs for review workflow clarity
Section titled “Structured hook outputs for review workflow clarity”Jan 19, 2026 · @mavam, @claude
The review command’s hook scripts now output structured control signals that make the workflow more explicit and reliable. The scope detection hook outputs Scope:, Diff:, and file list lines, while the directory creation hook outputs Review directory: with the path. This eliminates ambiguity about what commands to run and where to write findings.
The changes also add design principles to the hook documentation, clarifying that hooks should output all computed values rather than requiring the model to construct them. The review documentation now shows concrete examples of constructing diff commands from hook output, and the reviewer skill guidance emphasizes focusing on changed code rather than the entire codebase.
Automated review setup with project context awareness
Section titled “Automated review setup with project context awareness”Jan 19, 2026 · @mavam, @claude
The review command now features automated setup and project-aware reviews through hook-based automation and an Explore agent phase.
Two new hooks run automatically before spawning reviewers:
- Scope detection hook determines what to review (staged, unstaged, or branch changes) and outputs the file list
- Review directory hook creates a timestamped directory at
.reviews/<date>/<session_id>/for persistent findings
Before spawning specialized reviewers, the command launches an Explore agent to gather project context by examining config files and the README. This context calibrates reviewer relevance scoring, with reviewers deducting confidence points when concerns don’t apply to the project.
The reviewing methodology now emphasizes finding relevance to the specific project context, helping filter out generic concerns that don’t match the codebase’s nature (e.g., performance concerns for static documentation sites).
Category indicators and refined reviewer specializations
Section titled “Category indicators and refined reviewer specializations”Jan 19, 2026 · @mavam, @claude
Code review findings now include visual category indicators and refined reviewer specializations for clearer prioritization.
The review command splits the previous consistency reviewer into two focused reviewers:
- Readability (👁️) evaluates naming quality, idiomatic patterns, and code clarity
- Performance (🚀) assesses algorithmic complexity, memory usage, and resource efficiency
Review output now displays category emojis alongside findings to quickly identify the concern area:
- 🛡️ security
- 🏗️ arch
- 🧪 tests
- 🎨 ux
- 👁️ readability
- 📖 docs
- 🚀 perf
This makes it easier to scan review results and understand both the urgency (from action emojis) and the nature (from category emojis) of each finding at a glance.
Command-scoped documentation synchronization hook
Section titled “Command-scoped documentation synchronization hook”Jan 17, 2026 · @mavam, @claude
The documentation synchronization hook now runs only during the /docs:write command instead of on every file operation. The hook configuration moved from hooks/hooks.json to the command’s frontmatter, making the behavior more explicit and reducing unnecessary synchronization checks.
🐞 Bug Fixes
Section titled “🐞 Bug Fixes”Worktree isolation for documentation repository clones
Section titled “Worktree isolation for documentation repository clones”Jan 21, 2026 · @mavam, @claude
The docs plugin now properly isolates .docs/ clones across git worktrees, preventing the docs:writer agent from discovering and using another worktree’s .docs/ directory instead of cloning its own copy.
The fix includes three improvements:
- Init hook runs on first tool call: Updated hook matchers in
write.md,review.md, andpr.mdto use*for the initialization hook, ensuring.docs/is cloned before any tool executes - Clear error messages on clone failure: The
synchronize-docs.shscript now blocks with an actionable message if the clone fails, helping users diagnose SSH key or network issues - Explicit isolation guidance: Added instructions to
writer.mdtelling the agent not to search for.docs/in other locations, reinforcing the one-clone-per-worktree design
This resolves scenarios where multiple worktrees would inadvertently share a single .docs/ clone, causing documentation changes to appear in unexpected locations.
Correct PR creation syntax in finalize command
Section titled “Correct PR creation syntax in finalize command”Jan 19, 2026 · @mavam, @claude
The finalize command documentation now correctly instructs users to invoke /git:pr as a command rather than spawning @git:pr as a subagent. This fixes incorrect guidance that would have caused the command to fail when attempting to create a pull request.
Additionally, agent references @ship:adder and @git:committer no longer use code formatting for consistency with agent reference style elsewhere in the documentation.
Automatic branch switching after merged documentation branches
Section titled “Automatic branch switching after merged documentation branches”Jan 17, 2026 · @mavam, @claude
The .docs directory now automatically switches to main when a topic branch has been merged, preventing stale documentation from being used during editing. Previously, the synchronization hook would only warn about merged branches but leave the repository on the stale branch, causing docs:write to work with outdated content.