devtronic 1.2.4 → 1.2.6

package/templates/addons/auto-devtronic/agents/afk-task-validator.md

@@ -0,0 +1,78 @@
+---
+name: afk-task-validator
+description: Analyzes GitHub issues and task descriptions for AFK-readiness. Calculates viability score (0-100), detects quality gaps, and provides interactive guidance for refinement.
+tools: Bash, Read, Grep, Glob
+model: haiku
+---
+
+You are an expert task analyzer specializing in evaluating whether software engineering tasks are suitable for autonomous ("AFK" = Away From Keyboard) execution. Your job is to thoroughly analyze task descriptions, calculate a viability score, identify quality gaps, and guide users toward clarity.
+
+## When to Invoke
+
+Claude should invoke you when:
+- User invokes `/validate-task-afk <input>`
+- Need to analyze a GitHub issue or task description for AFK-readiness
+- Integration with `/auto-devtronic --validate` flag
+- User needs interactive guidance to refine unclear tasks
+
+---
+
+## Analysis Framework
+
+You analyze tasks across 5 dimensions, each contributing to a weighted viability score (0-100):
+
+### 1. Clarity (25% weight)
+**Purpose**: Acceptance criteria must be measurable and executable.
+
+**Scoring**:
+- **95-100**: Explicit measurable outcomes
+- **70-90**: Clear requirements with some detail
+- **40-70**: Vague descriptions
+- **0-40**: Completely unmeasurable
+
+### 2. Scope (25% weight)
+**Purpose**: Task should affect 1-4 files, single feature, not architectural.
+
+**Scoring**:
+- **90-100**: 1-2 files, single feature
+- **70-85**: 2-4 files, related changes
+- **40-70**: Mentions 5+ files or multiple features
+- **0-40**: Architectural keywords (refactor, migrate, rewrite, redesign)
+
+### 3. Precedent (20% weight)
+**Purpose**: Similar patterns should exist in codebase (reduces unknown unknowns).
+
+### 4. Coverage (20% weight)
+**Purpose**: Existing test coverage ensures VERIFY will catch failures.
+
+### 5. Dependencies (10% weight)
+**Purpose**: Task should be self-contained (not blocked by external PRs/issues).
+
+---
+
+## Weighted Score Calculation
+
+```
+Total Score = (Clarity x 0.25) + (Scope x 0.25) + (Precedent x 0.20) + (Coverage x 0.20) + (Dependencies x 0.10)
+
+Score Interpretation:
+- 70-100: AFK VIABLE - proceed with /auto-devtronic --afk
+- 40-70:  MEDIUM RISK - recommend HITL mode (/auto-devtronic --hitl)
+- 0-40:   NEEDS REFINEMENT - ask user clarifying questions first
+```
+
+---
+
+## Output Format
+
+Generate markdown report with score, dimension breakdown, gaps with suggestions, and recommendation (AFK / HITL / Needs Refinement).
+
+## Rules
+
+1. **Be thorough**: Check all 5 dimensions, don't skip any
+2. **Use grep for precedent**: Search codebase for similar patterns, don't guess
+3. **Flag ambiguity**: If unclear, ask; don't assume
+4. **Provide examples**: When suggesting templates, include realistic examples from the domain
+5. **Interactive refinement**: If score <70, engage user with questions to improve it
+6. **No rejection**: Never reject a task outright; always offer a path to improvement
+7. **Document assumptions**: If you can't run a tool (e.g., vitest), note it in the report

package/templates/addons/auto-devtronic/agents/{quality-runner.md → quality-executor.md}

@@ -1,5 +1,5 @@
 ---
-name: quality-runner
+name: quality-executor
 description: Runs all quality checks (typecheck, lint, test) for the auto-devtronic loop. Returns structured pass/fail output for failure-analyst to process.
 tools: Bash
 model: haiku

package/templates/addons/auto-devtronic/manifest.json

@@ -10,7 +10,8 @@
     "agents": [
       "issue-parser",
       "failure-analyst",
-      "quality-runner"
+      "quality-executor",
+      "afk-task-validator"
     ]
   }
 }

package/templates/addons/auto-devtronic/skills/auto-devtronic/SKILL.md

@@ -195,7 +195,7 @@ Comments: <comments>
 Output the brief in the format specified in your instructions.
 ```
 
-The agent outputs a `thoughts/devtronic/brief.md` file with the 3-layer structure:
+The agent outputs a `thoughts/auto-devtronic/brief.md` file with the 3-layer structure:
 
 ```markdown
 # Brief: <issue title>
@@ -255,11 +255,11 @@ the human review gate. Equivalent to running AFK mode for this one step only.
 
 Follow the same process as `/generate-tests`:
 
-Read `thoughts/devtronic/brief.md`, specifically the **Validation** layer, and generate failing tests that encode each acceptance criterion.
+Read `thoughts/auto-devtronic/brief.md`, specifically the **Validation** layer, and generate failing tests that encode each acceptance criterion.
 
 Save to:
 - Test files in the appropriate test directory (detect from project structure)
-- `thoughts/devtronic/tests-manifest.md` — traceability: criterion → test name → file
+- `thoughts/auto-devtronic/tests-manifest.md` — traceability: criterion → test name → file
 
 **Test naming convention**: `[feature].[criterion-slug].test.ts` or follow existing conventions.
 
@@ -293,10 +293,10 @@ Options:
 ## Step 4: Create Plan
 
 Follow the same process as `/create-plan`, using:
-- `thoughts/devtronic/brief.md` as the spec
-- `thoughts/devtronic/tests-manifest.md` as the DoD
+- `thoughts/auto-devtronic/brief.md` as the spec
+- `thoughts/auto-devtronic/tests-manifest.md` as the DoD
 
-Save plan to `thoughts/devtronic/plan.md`.
+Save plan to `thoughts/auto-devtronic/plan.md`.
 
 The plan **must include**:
 - `## Task Dependencies` YAML block (required for parallel execution)
@@ -314,7 +314,7 @@ This is the RALPH core. Runs until tests pass or retries are exhausted.
 attempt     = 1
 max_retries = N (from --max-retries, default 3)
 last_failure = null
-plan_path   = thoughts/devtronic/plan.md
+plan_path   = thoughts/auto-devtronic/plan.md
 ```
 
 ### Loop body
@@ -327,7 +327,7 @@ WHILE attempt <= max_retries:
   Include last_failure context in subagent prompts (if attempt > 1).
 
   ── 5b. Run Quality Checks ──────────────────────────────
-  Invoke the quality-runner agent:
+  Invoke the quality-executor agent:
   "Run all quality checks (typecheck, lint, test)"
 
   ── 5c. Evaluate Result ─────────────────────────────────
@@ -337,7 +337,7 @@ WHILE attempt <= max_retries:
   IF checks fail:
     last_failure = {
       attempt: N,
-      output: <quality-runner output>,
+      output: <quality-executor output>,
       affected_files: <list from git diff>
     }
 
@@ -376,7 +376,7 @@ When retries are exhausted or analyst recommends escalating:
 ## devtronic: Cannot Self-Correct ⚠️
 
 **Attempts**: N / max_retries
-**Last failure**: [quality-runner output summary]
+**Last failure**: [quality-executor output summary]
 
 **Failure analysis**:
 - Root cause: [description]
@@ -536,9 +536,9 @@ git stash list | grep "devtronic" && git stash pop
 | `path/to/file.ts` | Added / Modified |
 
 ### Session artifacts
-- Brief: `thoughts/devtronic/brief.md`
-- Tests manifest: `thoughts/devtronic/tests-manifest.md`
-- Plan: `thoughts/devtronic/plan.md`
+- Brief: `thoughts/auto-devtronic/brief.md`
+- Tests manifest: `thoughts/auto-devtronic/tests-manifest.md`
+- Plan: `thoughts/auto-devtronic/plan.md`
 ```
 
 ---
@@ -608,4 +608,4 @@ If a subagent task fails due to conflicts:
 - **Requires devtronic core skills** — devtronic uses the same logic as `/generate-tests`, `/create-plan`, and `/execute-plan`. Those skills must be installed.
 - **Requires `gh` CLI** — authenticated with repo write access.
 - **AFK is not magic** — it reduces friction, not mistakes. Start with HITL on unfamiliar codebases.
-- **Session artifacts** — all intermediate files are saved to `thoughts/devtronic/`. Do not delete them until the PR is merged.
+- **Session artifacts** — all intermediate files are saved to `thoughts/auto-devtronic/`. Do not delete them until the PR is merged.

package/templates/addons/design-best-practices/manifest.json

@@ -7,9 +7,9 @@
   "files": {
     "skills": [
       "design-init",
-      "design-review",
+      "design-critique",
       "design-refine",
-      "design-system",
+      "design-tokens",
       "design-harden"
     ],
     "reference": [

package/templates/addons/design-best-practices/skills/{design-review → design-critique}/SKILL.md

@@ -1,7 +1,8 @@
 ---
-name: design-review
+name: design-critique
 description: Design critique with AI slop detection — visual hierarchy, information architecture, emotional resonance, actionable report
 user-invokable: true
+allowed-tools: Read, Glob, Grep
 ---
 
 # Design Review

package/templates/addons/design-best-practices/skills/design-harden/SKILL.md

@@ -2,6 +2,7 @@
 name: design-harden
 description: Production hardening — text overflow, i18n, error states, accessibility, responsive, edge cases with severity report
 user-invokable: true
+allowed-tools: Read, Glob, Grep
 ---
 
 # Design Harden

package/templates/addons/design-best-practices/skills/design-init/SKILL.md

@@ -2,6 +2,7 @@
 name: design-init
 description: One-time project design context setup — gathers visual direction, constraints, and writes Design Context to CLAUDE.md
 user-invokable: true
+allowed-tools: Read, Write, Edit, Glob, Grep, AskUserQuestion
 ---
 
 # Design Init

package/templates/addons/design-best-practices/skills/design-refine/SKILL.md

@@ -2,6 +2,8 @@
 name: design-refine
 description: Directional design refinement — push the design bolder, quieter, more minimal, or more delightful
 user-invokable: true
+allowed-tools: Read, Edit, Glob, Grep, AskUserQuestion
+argument-hint: "--direction [bolder|quieter|minimal|delightful]"
 args:
   - name: direction
     description: "Refinement direction: bolder, quieter, minimal, delightful"

package/templates/addons/design-best-practices/skills/{design-system → design-tokens}/SKILL.md

@@ -1,7 +1,9 @@
 ---
-name: design-system
+name: design-tokens
 description: Design system extraction and normalization — find patterns or apply consistent tokens across the codebase
 user-invokable: true
+allowed-tools: Read, Edit, Glob, Grep
+argument-hint: "--mode [extract|normalize]"
 args:
   - name: mode
     description: "Mode: extract (find patterns and tokens) or normalize (apply system consistently)"

package/templates/claude-code/.claude/rules/quality.md

@@ -1,3 +1,7 @@
+---
+alwaysApply: true
+---
+
 # Code Quality Checks
 
 After implementing a new feature or modifying existing code, **always run the following checks** before considering the task complete.

package/templates/claude-code/.claude/skills/audit/SKILL.md

@@ -1,5 +1,5 @@
 ---
-name: devtronic:audit
+name: audit
 description: "Comprehensive codebase audit (SonarQube lite): architecture, code smells, complexity, security, coverage, and technical debt."
 allowed-tools: Read, Grep, Glob, Bash, Task, Write, AskUserQuestion
 argument-hint: "[--architecture|--code|--complexity|--security|--quick|directory]"

package/templates/claude-code/.claude/skills/backlog/SKILL.md

@@ -1,5 +1,5 @@
 ---
-name: devtronic:backlog
+name: backlog
 description: Manage the issue backlog. Add, prioritize, start work, complete, and cleanup items. Keeps the file manageable with automatic limits.
 allowed-tools: Read, Write, Edit, Glob, AskUserQuestion
 argument-hint: "[add|move|cleanup] [args]"

package/templates/claude-code/.claude/skills/brief/SKILL.md

@@ -1,5 +1,5 @@
 ---
-name: devtronic:brief
+name: brief
 description: Quick contextual briefing with pre-flight validation. Scans docs, code, git, and runs health checks before starting work.
 allowed-tools: Task, Glob, Grep, Read, Bash
 argument-hint: "[topic]"

package/templates/claude-code/.claude/skills/briefing/SKILL.md

@@ -1,5 +1,5 @@
 ---
-name: devtronic:briefing
+name: briefing
 description: Pre-planning alignment Q&A. Generates targeted questions to clarify scope, style, priority and constraints before diving into implementation.
 allowed-tools: Read, Write, Bash, Glob, Grep, AskUserQuestion
 argument-hint: "[topic]"

package/templates/claude-code/.claude/skills/checkpoint/SKILL.md

@@ -1,5 +1,5 @@
 ---
-name: devtronic:checkpoint
+name: checkpoint
 description: Save current session progress for resumption. Creates a compact state document with what's done, what's pending, and how to continue.
 allowed-tools: Read, Write, Bash, Glob
 argument-hint: "[task-name]"

package/templates/claude-code/.claude/skills/create-plan/SKILL.md

@@ -1,5 +1,5 @@
 ---
-name: devtronic:create-plan
+name: create-plan
 description: Strategic planning for complex tasks. Default creates phased implementation plan. Use --detailed for task-level pseudocode breakdown.
 allowed-tools: Read, Grep, Glob, Bash, AskUserQuestion, Write
 argument-hint: "[feature] [--detailed|--from-spec]"

package/templates/claude-code/.claude/skills/create-skill/SKILL.md

@@ -1,5 +1,5 @@
 ---
-name: devtronic:create-skill
+name: create-skill
 description: Meta skill to create new skills. Guides through defining purpose, workflow steps, tools, and generates the skill file.
 allowed-tools: AskUserQuestion, Write, Read, Glob
 argument-hint: "[skill-name]"

package/templates/claude-code/.claude/skills/design/SKILL.md

@@ -1,5 +1,5 @@
 ---
-name: devtronic:design
+name: design
 description: Orchestrates the design phase. Detects context and delegates to the right sub-skill.
 allowed-tools: Task, Read, Glob, AskUserQuestion, Write
 argument-hint: "[--research|--define|--ia|--wireframe|--system|--audit|--review|--spec]"

package/templates/claude-code/.claude/skills/design-audit/SKILL.md

@@ -1,5 +1,5 @@
 ---
-name: devtronic:design-audit
+name: design-audit
 description: UX audit — evaluates designs against Nielsen's 10 heuristics and WCAG 2.1 AA accessibility. Consolidates findings with severity.
 allowed-tools: Task, Read, Glob, Write, Bash
 argument-hint: "[--wireframes|--code|--both]"

package/templates/claude-code/.claude/skills/design-define/SKILL.md

@@ -1,5 +1,5 @@
 ---
-name: devtronic:design-define
+name: design-define
 description: Synthesis phase — generates personas, user journey maps, and HMW problem statements from research.
 allowed-tools: Task, Read, Write, AskUserQuestion
 argument-hint: "[topic]"

package/templates/claude-code/.claude/skills/design-ia/SKILL.md

@@ -1,5 +1,5 @@
 ---
-name: devtronic:design-ia
+name: design-ia
 description: Information architecture phase — defines navigation structure, sitemap, and user flows from personas and journeys.
 allowed-tools: Task, Read, Write, AskUserQuestion
 argument-hint: "[scope]"

package/templates/claude-code/.claude/skills/design-research/SKILL.md

@@ -1,5 +1,5 @@
 ---
-name: devtronic:design-research
+name: design-research
 description: Discovery phase — synthesizes competitive analysis, target audience, and business context into a research document.
 allowed-tools: Task, Read, Glob, AskUserQuestion, Write, WebFetch
 argument-hint: "[topic]"

package/templates/claude-code/.claude/skills/design-review/SKILL.md

@@ -1,5 +1,5 @@
 ---
-name: devtronic:design-review
+name: design-review
 description: QA visual review — compares implementation against wireframe specs and design system. Surfaces deviations with severity.
 allowed-tools: Task, Read, Glob, Bash, Write
 argument-hint: "[component-name|--all]"

package/templates/claude-code/.claude/skills/design-spec/SKILL.md

@@ -1,5 +1,5 @@
 ---
-name: devtronic:design-spec
+name: design-spec
 description: Generates a developer-ready design specification from design artifacts. Bridges design phase and implementation planning.
 allowed-tools: Task, Read, Write, Glob
 argument-hint: "[screen-name|--all]"

package/templates/claude-code/.claude/skills/design-system/SKILL.md

@@ -1,5 +1,5 @@
 ---
-name: devtronic:design-system
+name: design-system
 description: Design system router — delegates to define, audit, or sync based on flag. Manages design tokens, components, and style guide lifecycle.
 allowed-tools: Task, AskUserQuestion
 argument-hint: "[--define|--audit|--sync]"

package/templates/claude-code/.claude/skills/design-system-audit/SKILL.md

@@ -1,5 +1,5 @@
 ---
-name: devtronic:design-system-audit
+name: design-system-audit
 description: Audits the codebase for design system drift — finds hardcoded values, unused tokens, and components that bypass the system.
 allowed-tools: Task, Read, Glob, Grep, Write, Bash
 argument-hint: "[--files path/to/dir]"

package/templates/claude-code/.claude/skills/design-system-define/SKILL.md

@@ -1,5 +1,5 @@
 ---
-name: devtronic:design-system-define
+name: design-system-define
 description: Creates or extracts a design system — tokens, components, and style guide — persisted to thoughts/design/design-system.md.
 allowed-tools: Task, Read, Write, Edit, AskUserQuestion, Bash, Glob
 argument-hint: "[--extract]"

package/templates/claude-code/.claude/skills/design-system-sync/SKILL.md

@@ -1,5 +1,5 @@
 ---
-name: devtronic:design-system-sync
+name: design-system-sync
 description: Syncs design tokens from thoughts/design/design-system.md to project config files (Tailwind, CSS vars, tokens.json).
 allowed-tools: Task, Read, Write, Edit, Glob, Bash, AskUserQuestion
 argument-hint: "[--dry-run]"

package/templates/claude-code/.claude/skills/design-wireframe/SKILL.md

@@ -1,5 +1,5 @@
 ---
-name: devtronic:design-wireframe
+name: design-wireframe
 description: Wireframe specification phase — generates structured text-based wireframe specs per screen. Tool-agnostic. Reads IA and define artifacts.
 allowed-tools: Task, Read, Write, AskUserQuestion
 argument-hint: "[screen-name|--all]"