agentify v0.1.1

MVP-hardening release since v0.1.0.

The deliverable is now a real tool. agentify assemble turns a markdown design into the self-contained, themed design.html (embedded diagrams, em-dash guard). The full render-then-assemble flow is verified to run from the extracted skill package with no repo access.

The reasoning is under test. A new eval harness (eval/cases.jsonl + eval/check.mjs, run in CI) grades a produced design against expected decisions. Five committed gallery designs are seeded and passing.

The knowledge base was audited against primary sources (nearly all load-bearing claims verified exactly; three corrected).

New reasoning rules: domain-harm analysis (scale guardrails to harm, not company size) and right-sizing (lightweight vs enterprise, so small designs are not buried in enterprise ceremony).

Voice and multimodal are now grounded (cascaded vs speech-to-speech, latency budget, turn-taking, audio-native eval), with asr/tts diagram nodes.

Five gallery examples: enterprise support agent, legal document assistant, autonomous coding system (interview mode), lightweight recipe assistant, and a drive-through voice agent.

Install: npx skills add avnath13/agentify -g

agentify v0.1.0

First release: an enterprise agentic system design skill.

• Grounded knowledge base (14 documents) synthesized into 7 decision trees
• Enterprise AI solutions architect persona, production and interview modes
• Diagram engine extended with 11 agent-native component types
• Flagship gallery example: a full enterprise support agent design document
• Install-free distributable; tests and style lint green

Install: npx skills add avnath13/agentify -g