v2.1.4 — Decision Supersession Tracking & Traceability

What's New

Decision supersession tracking (PR #2)

• Added Supersedes: DEC-NNN field to decision log for /reopen and /alternative commands
• Added References: FR-NNN, DEC-NNN field for cross-phase traceability
• Added Traceability section (section 9) to /design-component output with requirement checklists
• Updated state-manager skill with supersession and traceability guidance and examples
• Updated /refine, /accept, /generate-docs with traceability support

Improvements

• Mapped workflow commands to phases in README Commands section
• Renamed Tessl tiles to emerge/ workspace, set private: false

Full Changelog: v2.1.3...v2.1.4

v2.1.3 — Tessl Integration & Post-Merge Fixes

Tessl integration (PR #1)

• Added Tessl tile packaging for all 4 architecture skills
• Added 20 evaluation scenarios (5 per skill)
• Expanded skill SKILL.md files with activation triggers and detailed instructions
• Added CONTRIBUTING.md with Tessl integration guide
• Updated CLAUDE.md with Phase 2A/2B/2C sub-phase rules and reopen limits

Post-merge fixes

• Restored emphatic "ONLY" casing in state-manager SKILL.md
• Added integrity placeholder to tessl.json dependency
• Fixed stale SVG diagram references in CONTRIBUTING.md
• Fixed duplicate METHODOLOGY.md entry in CONTRIBUTING.md project tree

Full Changelog: v2.1.2...v2.1.3

v2.1.2 — Documentation Overhaul

Documentation overhaul

• Converted all diagrams across the repository to Mermaid syntax (README, ARCHITECTURE.md, METHODOLOGY.md)
• Added end-to-end workflow explanation with step-by-step journey
• Added artifact flow diagram showing how architecture knowledge builds across phases
• Added full e-commerce platform example scenario with per-phase artifacts
• Added generated artifacts table with locations and timing
• Added use case positioning table
• Updated system-level diagram to show actual file paths
• Replaced non-existent SVG references in ARCHITECTURE.md with Mermaid diagrams
• Converted component lifecycle ASCII art to Mermaid state diagrams

Full Changelog: v2.1.0...v2.1.2

v2.1.0 — Import Existing Architecture

What's New

Import Existing Architecture

Have an existing architecture document? Import it for review and iteration:

npx arch-agent import path/to/architecture.md --name "My Project"

Then in Claude Code:

/import-architecture

The agent parses your document and walks through each phase — extracting requirements, architecture pattern, component map, cross-cutting decisions, and individual component designs. Each phase gets adversarial review with standard acceptance gates.

Features

• New CLI command: arch-agent import <source> — scaffolds project with imported document
• New slash command: /import-architecture — fast-track review flow
• Higher reopen limit: Imported projects get 5 reopens (vs 2) for more iteration room
• Zero validator changes: Uses the exact same state machine as normal flow
• 9 new import tests + 2 CLI tests (51 total)

Install

npx arch-agent import existing-architecture.md --name "My Project"

v2.0.1 — Security Hardening + npm as arch-agent

What's New

Package renamed to arch-agent

npm rejected "architor" (too similar to "archiver"). The package is now published as arch-agent:

npx arch-agent init --name "My Project"

Security Hardening (9 zero-impact fixes)

All fixes block adversarial/malformed input only. Normal workflow is 100% identical.

Critical

• Closed Write+Execute chain: scripts are now read-only

High

• Reopen limit bypass: reopens.max read from trusted on-disk state
• Unknown phase bypass: phase name allowlist blocks unrecognized phases
• Dynamic all_accepted: finalization gate computes from actual component statuses

Medium

• Schema type validation enforced
• Log sanitization for markdown-active characters

Low

• stdin limited to 1MB in validation hook
• Backup failures logged to stderr

Other

• Security Constraints section added to CLAUDE.md
• Symlink safety in CLI walkDir
• 10 new security tests (40 total)

Install

npx arch-agent init