Claude Code Plugins

Community-maintained marketplace

Feedback
11
0

Reviews Stories/Tasks against 2025 standards before approval (Backlog -> Todo). Auto-fixes issues, validates structure, optimizes via YAGNI/KISS/SOLID. Auto-discovers team/config.

Install Skill

1Download skill
2Enable skills in Claude

Open claude.ai/settings/capabilities and find the "Skills" section

3Upload to Claude

Click "Upload skill" and select the downloaded ZIP file

Note: Please verify skill by going through its instructions before using it.

SKILL.md

name ln-320-story-validator
description Reviews Stories/Tasks against 2025 standards before approval (Backlog -> Todo). Auto-fixes issues, validates structure, optimizes via YAGNI/KISS/SOLID. Auto-discovers team/config.

Story Verification Skill

Critically review and auto-fix Stories and Tasks against 2025 standards and project architecture before execution.

Purpose & Scope

  • Validate Story plus child Tasks against industry standards and project patterns
  • Auto-fix detected issues (structure, solution, workflow, quality) before approval
  • Approve Story after fixes (Backlog -> Todo) and link any missing guides
  • Strip or re-home Non-Functional Requirements; keep only functional scope in Story/Tasks

When to Use

  • Reviewing Stories before approval (Backlog -> Todo)
  • Validating implementation path across Story and Tasks
  • Ensuring standards, architecture, and solution fit
  • Optimizing or correcting proposed approaches

Research Tools (Mandatory)

For standards verification (Phase 2):

  • mcp__Ref__ref_search_documentation(query="[domain] RFC standard best practices 2025") - industry standards
  • mcp__Ref__ref_search_documentation(query="[library] security vulnerabilities OWASP") - security checks
  • WebSearch(query="[library] latest stable version 2025") - version verification

For library versions (Auto-Fix #6):

  • mcp__context7__resolve-library-id(libraryName="[library]") → get Context7 ID
  • mcp__context7__get-library-docs(context7CompatibleLibraryID="...", topic="version changelog") - current versions

Time-box: 5-10 minutes per Story for research; skip if Story is trivial CRUD

Workflow Overview

Phase 1: Discovery & Loading

  • Auto-discover configuration: Team ID (docs/tasks/kanban_board.md), project docs (CLAUDE.md), epic from Story.project
  • Load metadata only: Story ID/title/status/labels, child Task IDs/titles/status/labels
  • Expect 3-8 implementation tasks; record parentId for filtering
  • Rationale: keep loading light; full descriptions arrive in Phase 2

Phase 2: Critical Solution Review

  • Load full Story description (all 8 sections) when analysis starts
  • Standards first (priority): Industry/RFCs -> Security -> 2025 best practices -> KISS/YAGNI/DRY within standards
  • Research using MCP Ref: call ref_search_documentation(query="[story domain] RFC standard 2025") for each technology in Story
  • Verify versions via Context7: call get-library-docs to check library versions mentioned in Technical Notes
  • Challenge approach: prefer proven standards over custom work; cite specific RFC/standard in Linear comment when fixing
  • Documentation check and creation triggers (pattern keywords, package versions, technology choices); invoke ln-321-best-practices-researcher(doc_type="guide/manual/adr", topic="...") if missing
  • Verify against codebase reality before edits; if Story is outdated, auto-correct via Linear update
  • Reporting rule: when principles are violated, explain why with best-practice references and propose concrete fixes, not just list names

Phase 3: Comprehensive Auto-Fix

  • Always auto-fix; no "Needs Work" path. Follow execution order: Structural (1-4) -> Solution (5-8) -> Workflow (9-12) -> Scope & Quality (13-16)
  • Use Auto-Fix Actions table below as the authoritative checklist; keep sequential task validation to avoid truncation
  • Test mention: ensure Test Strategy section exists but keep it empty here; do not plan coverage or execution at this stage

Phase 4: Approve & Notify

  • Set Story + all Tasks to Todo (Linear); update kanban_board.md with APPROVED marker and move items from Backlog to Todo keeping hierarchy
  • Add Linear comment summarizing fixes, created docs (paths), ADRs/manuals/guides, and any TODO warnings
  • Verification summary display: Story verdict (always Approved -> Todo), changes, linked docs, warnings; Tasks table with changes and guide links
  • Any principle violations in reports must include rationale, best-practice reference, and recommended remediation steps

Auto-Fix Actions Reference

# What it checks Auto-fix actions Notes/Evidence
1 Story Structure 8 sections per story_template_universal.md in order Add/reorder sections and subsections with TODO placeholders; update Linear; preserve language Skip if Done/Canceled or older than 30 days
2 Tasks Structure Each Task has 7 sections in order Load each Task full description sequentially; add/reorder sections with placeholders; update Linear and comment; preserve language Sequential per task; skip if Done/Canceled or older than 30 days
3 Story Statement As a/I want/So that clarity Rewrite using persona (Context), capability (Technical Notes), value (Success Metrics); update Linear and comment -
4 Acceptance Criteria Given/When/Then, 3-5 items, edge/error coverage Normalize AC to G/W/T; add missing edge/error cases; update Linear and comment -
5 Solution Optimization 2025-best approach via MCP Ref research Call ref_search_documentation(query="[domain] best practices 2025"); rewrite Technical Notes if better approach found; cite RFC/standard number; add Linear comment with research evidence Evidence: MCP Ref query + result summary
6 Library & Version Verify via Context7 + WebSearch Call mcp__context7__get-library-docs(topic="version") for each library; if outdated, update to latest stable; log checked versions in Linear comment Evidence: Context7 query + version comparison
7 Test Strategy Section exists but remains empty now Ensure Test Strategy section present; leave empty with note that testing is planned later by ln-350; do not add coverage details Mention "testing handled later; not evaluated in this phase"
8 Documentation Integration No standalone doc tasks Remove doc-only tasks; fold doc updates into implementation tasks and DoD; update Linear and comment -
9 Story Size & Granularity 3-8 tasks; 3-5h each If <3 tasks invoke ln-310-story-decomposer; if >8 add TODO to split; flag tasks <3h or >8h with TODO; reload metadata and update kanban_board.md Comment creation source
10 Test Task Cleanup No premature test tasks Cancel/remove tasks labeled as tests or named test/comprehensive/final; remove from Story/kanban; comment that testing tasks appear later Testing not executed now
11 YAGNI No premature features Move speculative items to Out of Scope/Future; update Technical Notes/Tasks; comment rationale -
12 KISS Simplest solution within standards Replace over-engineered parts with simpler options unless standards require otherwise; update Linear with reasoning Standard overrides simplicity
13 Documentation Links Technical Notes reference docs Add "Related Documentation" subsection; link guides/manuals/ADRs by path; update Linear Use created/existing paths
14 Foundation-First Order Task order DB -> Repo -> Service -> API -> Frontend Reorder Implementation Tasks and note execution order; update Linear and comment -
15 Code Quality Basics No magic values; config approach defined Add TODOs for constants/config/env creds; describe config management in Technical Notes; update Linear Warn in summary if TODOs remain
16 Industry Standards MCP Ref verification mandatory Call ref_search_documentation(query="[domain] RFC OWASP standard") BEFORE checking local docs; if RFC applies, verify Story complies; if local doc missing, invoke creators; add Linear comment with RFC reference or explicit "no standards apply" with reasoning Evidence: MCP Ref query result or explicit skip reason
17 Rate Limiting API Stories must have Rate Limiting subsection If API Story (keywords: endpoint, API, REST, GraphQL): check Rate Limiting in Technical Notes; if missing, add placeholder with TODO (Policy, Strategy, Headers, 429 Response); update Linear Skip if internal service, no external API
18 Auth/Security Pattern Auth Stories must have Auth subsection If Auth Story (keywords: auth, login, token, JWT, OAuth): check Authentication & Authorization in Technical Notes; if missing or generic, add TODO for token management, authorization model; update Linear Skip if non-auth Story
19 Error Handling Strategy Story + all Tasks must have Error Handling Check Story Technical Notes for Error Handling Strategy; check each Task for Error Handling Strategy section; add TODO placeholders if missing (error format, codes, retry, circuit breaker); update Linear Mandatory for API Stories
20 Logging & Observability Story + all Tasks must have Logging Check Story Technical Notes for Logging & Observability; check each Task for Logging Requirements section; add TODO placeholders if missing (log levels, format, audit trail, tracing); update Linear Mandatory for all Stories

Self-Audit Protocol (Mandatory)

  • Before marking any criterion, answer its Self-Audit question, provide concrete evidence (doc path, Linear comment link, worker result), and record in checkpoints/{STORY_ID}_verification_log.md. No evidence -> no completion.
# Self-Audit Question Required Evidence
1 Validated all 8 Story sections in order? Section list
2 Loaded full description for each Task? Task validation count
3 Statement in As a/I want/So that? Quoted statement
4 AC are G/W/T and testable? AC count and format
5 Challenged "best for 2025" via MCP Ref? MCP Ref query + result summary
6 Versions verified via Context7? Context7 query + version list
7 Test Strategy kept empty for now? Note that testing is deferred
8 Docs integrated, no standalone tasks? Integration evidence
9 Task count 3-8 and 3-5h? Task count/sizes
10 No premature test tasks? Search result
11 Only current-scope features? Scope review
12 Simplest approach within standards? Simplicity justification
13 All relevant guides linked? Guide paths
14 Tasks ordered DB->Repo->Service->API? Task order list
15 Hardcoded values handled? TODO/config evidence
16 Standards verified via MCP Ref? MCP Ref query or explicit skip reason
17 Rate Limiting documented for API Story? Rate Limiting subsection or skip reason
18 Auth/Security pattern fully documented? Auth subsection or skip reason
19 Error Handling Strategy in Story AND all Tasks? Evidence path for each
20 Logging requirements in Story AND all Tasks? Evidence path for each

Definition of Done

  • Verification log created from template with answered Self-Audit for all 20 criteria and evidence
  • Phase 1: auto-discovery done; Story + Tasks metadata loaded; task count checked
  • Phase 2: full Story parsed; standards researched; doc gaps filled via creators; codebase reality verified; reports explain violations with best-practice references and fixes
  • Phase 3: criteria 1-20 auto-fixed in order; Test Strategy section present but empty; test tasks removed; guide links inserted; API Technical Aspects validated (#17-#20)
  • Phase 4: Story/Tasks set to Todo; kanban_board.md updated with APPROVED marker and hierarchy; Linear comment added with fixes, docs, ADRs/manuals/guides, TODO warnings; summary table shown

Example Workflows

  • Outdated library: Detect old version, update to current stable in Technical Notes/Tasks, document change, approve.
  • OAuth violation: Replace custom endpoints with RFC-compliant /token flow, add spec reference, update Tasks/Technical Notes, approve.
  • Missing tasks: If <3 tasks, run ln-310-story-decomposer, reload tasks, reorder Foundation-First, update kanban and comments, approve.

Reference Files

  • Templates: ../ln-220-story-coordinator/references/story_template_universal.md, ../ln-311-task-creator/references/task_template_implementation.md
  • Checklists: references/verification_checklist.md, references/verification_log_template.md
  • Testing methodology (for later phases): ../ln-350-story-test-planner/references/risk_based_testing_guide.md
  • Linear integration: ../ln-210-epic-coordinator/references/linear_integration.md

Version: 13.0.0 (Added criteria #17-#20: Rate Limiting, Auth/Security, Error Handling, Logging & Observability) Last Updated: 2025-12-12