v5.0.0 — Trust-Grade Framework Release by emredursun · Pull Request #16 · devran-ai/kit

emredursun · 2026-03-28T18:25:47Z

Summary

Major release: Devran AI Kit v5.0.0 — Trust-Grade AI Development Framework.

Framework Enrichment (Phase 1-8):

9 critical workflows enriched with Scope Filters, Failure Templates, Ethics Gates, Argument Parsing
6 medium workflows enriched (create, enhance, brainstorm, review, pr-review, ui-ux-pro-max)
New /implement workflow created (was missing despite command existing)
All 37 commands fully wired with workflow:, invokes:, uses: frontmatter declarations
All 23 workflow Output Templates given emoji-prefixed headers (📋🐛🚀🔍🧪 etc.)
8 new skill sub-files: pr-toolkit (3), verification-loop, production-readiness, intelligent-routing, security-practices, testing-patterns
New research-methodology skill (skills: 35 → 36)

New Governance Rules:

performance.md — JS bundle ≤200KB, API p95 ≤300ms, LCP ≤2.5s
accessibility.md — WCAG 2.1 AA, semantic HTML, color contrast 4.5:1
data-privacy.md — PII handling, GDPR, AI pipeline anonymization
All 3 wired into loading engine + planner Rule Consultation table

ECC-Inspired Innovations:

Instinct System (contexts/instincts.md) — confidence-scored pattern memory
Rigor Profiles (strict/standard/minimal) in workflow-standards
Session-context bootstrap fix (session-start hook was blocking)

OG Image Redesign (Vercel-grade):

Removed dynamic numbers (were going stale on every change)
Added "TRUST-GRADE AI DEVELOPMENT FRAMEWORK" subtitle
Monochrome Lucide SVG icons (cross-platform consistent)
Unified pill design language across all rows
Retina 2400×1260 @2x via Puppeteer generator (scripts/generate-og-image.js)

Badge System Overhaul:

3x larger badges (style=for-the-badge + height="36")
2-row layout: project meta + capabilities
Purple/violet brand family palette (harmonized from 7-color rainbow)
Auto-sync from manifest.json via sync-version.js
5 new structural tests for badge↔manifest drift detection

Infrastructure:

Windows EPERM retry logic for kit init file locks
sync-version.js: syncs version + all 5 capability badge counts from manifest
113 files changed, +4578 / -583 lines
573 tests passing (was 533)

Test plan

npm test — 573/573 passing
kit verify — manifest integrity check
Version sync: package.json ↔ manifest ↔ README badges ↔ architecture.md
All 23 workflow Output Templates have emoji headers
All 37 commands have frontmatter dependency declarations
OG image renders correctly at 1200×630 (Puppeteer verified)
Visual review of README on GitHub (OG image + badges)
npm publish --access public after merge

🤖 Generated with Claude Code

Root cause: fs.renameSync fails intermittently on Windows when antivirus, indexer, or vitest watcher holds a lock on the temp directory during atomic copy. Fix: - bin/kit.js: retry renameSync up to 3 times with 500ms backoff - tests/unit/cli.test.js: retry rmSync cleanup with force flag 568/568 tests passing (3 consecutive runs, 0 flakes).

- Replace busy-wait with Atomics.wait in kit init retry (gemini F3) - Re-throw error on final retry in test cleanup (gemini F4) - Add transient error code gating (EPERM/EACCES/EBUSY) to kit init - Extract TRANSIENT_FS_ERRORS constant to lib/constants.js (DRY) - Update lib/io.js and lib/ide-generator.js to use shared constant - Add user-visible retry warning log in kit init - Remove fragile shell sleep + require() in ESM test file 568/568 tests passing.

- Use fixed 500ms delay instead of linear backoff in kit init retry - Hoist rmWithRetry helper to describe scope for reuse across tests - Apply rmWithRetry to heal and exists tests (previously unprotected) Addresses gemini-code-assist review on commit a07eed0. 568/568 tests passing.

Only retry EPERM/EACCES/EBUSY in test cleanup helper — non-transient errors now fail fast instead of being masked by retry delays. Addresses gemini-code-assist review on commit 7fe3ded.

Pre-release commit: all workflows, commands, rules, skills, agents, checklists, templates, and documentation updates for v5.0.0.

Major release: Trust-Grade AI Development Framework v5.0.0 - 9 critical workflows enriched (Scope Filters, Failure Templates, Ethics gates) - 15 new files: 3 rules, 8 skill sub-files, 2 contexts, 1 workflow, 1 template - 37 commands fully wired with frontmatter dependency declarations - All 23 workflow Output Templates given emoji-prefixed headers - 36 skills, 23 workflows, 13 governance rules

…lity counts - Upgrade all shields.io badges from default to for-the-badge style (~2x larger) - Split badges into two rows: project meta (version/license/tests/deps) and capabilities (AI Agents / Skills / Commands / Workflows / Rules) - Add missing Commands, Workflows, Rules badges - sync-version.js: auto-sync all 5 capability counts from manifest.json on every `npm version` run — no manual badge updates needed - version-sync.test.js: add 5 structural tests to catch badge/manifest drift

@2x

OG Image: - Remove dynamic numbers (23 AGENTS, 34 SKILLS, etc.) that go stale - Replace with permanent feature highlights: Multi-Agent Orchestration, Trust-Grade Governance, Cross-IDE Support, Telegram Control - Add icon row: Agents, Skills, Commands, Workflows, Rules (no counts) - Retina-quality 2400x1260 @2x PNG via Puppeteer generator script - scripts/og-image.html: Source template (HTML/CSS) - scripts/generate-og-image.js: Automated generation script Badges: - Upgrade all shields.io badges to height="36" (~3x original size) - Two-row layout: project meta + capability counts - Update tests badge: 568 → 573 passing

OG Image (all P0-P3 fixes applied): - P0: System emoji → monochrome Lucide SVG icons (cross-platform) - P0: Add "TRUST-GRADE AI DEVELOPMENT FRAMEWORK" subtitle - P0: Restore original 3 channel labels (Claude Code, Telegram/Discord/iMessage, Remote Dev) - P1: Unify capability row with feature pill design language - P1: Optical center vertical balance (68px top / 52px bottom) - P2: Pill border opacity 0.08 → 0.12 for better contrast - P3: Glow line 2px → 3px with wider gradient spread Badge palette harmonization: - Row 2 capabilities: rainbow (7 colors) → purple/violet brand family AI Agents (#7c3aed), Skills (#8b5cf6), Commands (#6366f1), Workflows (#a78bfa), Rules (#9333ea) - sync-version.js: color-agnostic regex for future palette changes - version-sync.test.js: color-agnostic assertions

Replace space-between layout with centered single-block stack. All content flows as one cohesive unit with controlled gaps: Title (78px) → Subtitle (14px gap) → Tagline (22px) → Channel pills (30px) → Capability pills (24px). No more empty void between feature pills and bottom row.

- P0: Channel↔Capability gap 24px→18px (tighter grouping) - P1: Content nudge +6px down (title breathing room for social crop) - P2: Capability SVG stroke opacity 0.55→0.65 (readable on all screens) - P2: Capability label opacity 0.48→0.58 - P2: Capability pill border 0.07→0.09

gemini-code-assist

Code Review

This pull request upgrades the Devran AI Kit to version 5.0.0, implementing a comprehensive governance framework with new rules for performance, accessibility, and data privacy. Key additions include the /implement workflow, the research-methodology skill, and an 'Instinct System' for confidence-scored pattern memory. Review feedback suggests making dependencies explicit in the new OG image generation script and refactoring tests to use the centralized TRANSIENT_FS_ERRORS constant for better maintainability.

gemini-code-assist · 2026-03-28T18:28:17Z

scripts/generate-og-image.js

+  let puppeteer;
+  try {
+    puppeteer = require('puppeteer');
+  } catch {
+    console.error('❌ puppeteer not found. Installing...');
+    const { execSync } = require('child_process');
+    execSync('npm install --no-save puppeteer', { stdio: 'inherit' });
+    puppeteer = require('puppeteer');
+  }


This automatic installation of puppeteer is an anti-pattern. It creates a hidden dependency, has side effects, and won't work in environments without npm or write access.

The standard practice is to declare puppeteer as a devDependency in package.json and instruct developers to install it via npm install. This makes dependencies explicit and the project setup more predictable.

Please add puppeteer to your devDependencies and simplify this logic.

Suggested change

let puppeteer;

try {

puppeteer = require('puppeteer');

} catch {

console.error('❌ puppeteer not found. Installing...');

const { execSync } = require('child_process');

execSync('npm install --no-save puppeteer', { stdio: 'inherit' });

puppeteer = require('puppeteer');

}

const puppeteer = require('puppeteer');

gemini-code-assist · 2026-03-28T18:28:17Z

tests/unit/cli.test.js

+      }
+      return;
+    } catch (e) {
+      if (!['EPERM', 'EACCES', 'EBUSY'].includes(e.code) || i >= retries - 1) throw e;


For consistency and better maintainability, please use the TRANSIENT_FS_ERRORS constant that was introduced in lib/constants.js instead of hardcoding the array of error codes here.

This will ensure that if the list of transient errors is updated, it's changed in one central place.

You'll need to add an import at the top of the file:

import { TRANSIENT_FS_ERRORS } from '../../lib/constants.js';

And then update this line.

Suggested change

if (!['EPERM', 'EACCES', 'EBUSY'].includes(e.code) || i >= retries - 1) throw e;

if (!TRANSIENT_FS_ERRORS.has(e.code) || i >= retries - 1) throw e;

… constant - generate-og-image.js: Remove hidden `npm install --no-save puppeteer` auto-install. Fail with clear install instructions instead. - cli.test.js: Replace hardcoded ['EPERM','EACCES','EBUSY'] with TRANSIENT_FS_ERRORS constant from lib/constants.js (single source of truth)

… constant

emredursun · 2026-03-28T18:44:41Z

/gemini review

gemini-code-assist

Code Review

This pull request significantly upgrades the Devran AI Kit to version 5.0.0, introducing comprehensive workflow governance, new governance rules (Performance, Accessibility, Data Privacy), and an Instinct System for pattern memory. Key changes include adding Scope Filters, Failure Templates, and Argument Parsing tables to all workflows, implementing a new /implement workflow, and refining the documentation with new skill sub-files and improved command consistency. I have reviewed the provided comments and identified several formatting issues and naming inconsistencies that should be addressed to ensure a polished and consistent documentation set.

gemini-code-assist · 2026-03-28T18:47:54Z

.agent/skills/pr-toolkit/SKILL.md

+---
+
 ---


There appears to be a duplicate horizontal rule here. This is likely a formatting mistake and one of the --- lines should be removed for cleaner rendering.

Suggested change

---

---

---

gemini-code-assist · 2026-03-28T18:47:54Z

.agent/skills/security-practices/SKILL.md

+---
+
 ---


There appears to be a duplicate horizontal rule here. This is likely a formatting mistake and one of the --- lines should be removed for cleaner rendering.

Suggested change

---

---

---

gemini-code-assist · 2026-03-28T18:47:54Z

.agent/skills/testing-patterns/SKILL.md

+---
+
 ---


There appears to be a duplicate horizontal rule here. This is likely a formatting mistake and one of the --- lines should be removed for cleaner rendering.

Suggested change

---

---

---

gemini-code-assist · 2026-03-28T18:47:54Z

.agent/workflows/help-kit.md

+# /help-kit — Devran AI Kit Quick Reference

-> **Trigger**: `/help_kit`
+> **Trigger**: `/help-kit`


Throughout this pull request, command names are being standardized to use hyphens instead of underscores (e.g., /project-status). For consistency, this trigger should be updated from /help_kit to /help-kit.

Suggested change

> **Trigger**: `/help-kit`

> **Trigger**: /help-kit

gemini-code-assist · 2026-03-28T18:47:54Z

.agent/workflows/project-status.md

+# /project-status — Project Status Overview

-> **Trigger**: `/project_status [sub-command]`
+> **Trigger**: `/project-status [sub-command]`


The command name has been updated to /project-status in other files within this pull request. To maintain consistency across the documentation, this trigger should also be updated to use a hyphen instead of an underscore.

Suggested change

> **Trigger**: `/project-status [sub-command]`

> **Trigger**: /project-status [sub-command]

- Remove duplicate --- horizontal rules in 3 SKILL.md files (pr-toolkit, security-practices, testing-patterns) - Replace ALL remaining underscore command names with hyphens: /project_status → /project-status, /help_kit → /help-kit across CheatSheet.md, help.md, project-status.md, README.md, pr.md, upgrade.md, workflows/README.md (24 occurrences fixed) - Zero underscore command references remaining in .agent/

emredursun added 11 commits March 28, 2026 02:13

fix: gate rmWithRetry on transient error codes only

a0057d5

Only retry EPERM/EACCES/EBUSY in test cleanup helper — non-transient errors now fail fast instead of being masked by retry delays. Addresses gemini-code-assist review on commit 7fe3ded.

chore: stage all Phase 1-8 enhancements before v5.0.0 release

28ba8e6

Pre-release commit: all workflows, commands, rules, skills, agents, checklists, templates, and documentation updates for v5.0.0.

gemini-code-assist bot reviewed Mar 28, 2026

View reviewed changes

emredursun added 2 commits March 28, 2026 19:33

fix: resolve merge conflict in cli.test.js — keep TRANSIENT_FS_ERRORS…

797f9f9

… constant

gemini-code-assist bot reviewed Mar 28, 2026

View reviewed changes

emredursun merged commit f0cf02a into main Mar 28, 2026
4 checks passed

emredursun deleted the fix/windows-eperm-retry branch March 28, 2026 19:18

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

v5.0.0 — Trust-Grade Framework Release#16

v5.0.0 — Trust-Grade Framework Release#16
emredursun merged 14 commits intomainfrom
fix/windows-eperm-retry

emredursun commented Mar 28, 2026 •

edited

Loading

Uh oh!

gemini-code-assist bot left a comment

Uh oh!

gemini-code-assist bot Mar 28, 2026

Uh oh!

gemini-code-assist bot Mar 28, 2026

Uh oh!

emredursun commented Mar 28, 2026

Uh oh!

gemini-code-assist bot left a comment

Uh oh!

gemini-code-assist bot Mar 28, 2026

Uh oh!

gemini-code-assist bot Mar 28, 2026

Uh oh!

gemini-code-assist bot Mar 28, 2026

Uh oh!

gemini-code-assist bot Mar 28, 2026

Uh oh!

gemini-code-assist bot Mar 28, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

	if (!['EPERM', 'EACCES', 'EBUSY'].includes(e.code) \|\| i >= retries - 1) throw e;
	if (!TRANSIENT_FS_ERRORS.has(e.code) \|\| i >= retries - 1) throw e;

	> Trigger: `/project-status [sub-command]`
	> Trigger: /project-status [sub-command]

Conversation

emredursun commented Mar 28, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

Test plan

Uh oh!

gemini-code-assist bot left a comment

Choose a reason for hiding this comment

Code Review

Uh oh!

gemini-code-assist bot Mar 28, 2026

Choose a reason for hiding this comment

Uh oh!

gemini-code-assist bot Mar 28, 2026

Choose a reason for hiding this comment

Uh oh!

emredursun commented Mar 28, 2026

Uh oh!

gemini-code-assist bot left a comment

Choose a reason for hiding this comment

Code Review

Uh oh!

gemini-code-assist bot Mar 28, 2026

Choose a reason for hiding this comment

Uh oh!

gemini-code-assist bot Mar 28, 2026

Choose a reason for hiding this comment

Uh oh!

gemini-code-assist bot Mar 28, 2026

Choose a reason for hiding this comment

Uh oh!

gemini-code-assist bot Mar 28, 2026

Choose a reason for hiding this comment

Uh oh!

gemini-code-assist bot Mar 28, 2026

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

emredursun commented Mar 28, 2026 •

edited

Loading