sdlc-atlas

Plugin sdlc-atlas from marketplace trackmind in trackmind-ai/marketplace. Commands read from the repo's manifest — 22 days ago.

Skills (89)

• ado-setup

  One-time Azure DevOps integration setup. Reads or asks for org URL and project name (writes them to CLAUDE.md), checks that AZURE_DEVOPS_EXT_PAT is set as a User environment variable (never asks for the value), installs the ado-work-item skill, and verifies the connection. Called by /start.

• ado-work-item

  Fetch a work item from Azure DevOps by ID. Handles three shapes: (A) Task — also fetches parent Story for AC context; (B) User Story with no child tasks — returns story fields directly; (C) User Story with child Tasks — fetches and returns the full task list so /feature can present a selection promp

• agent-handoff

  Seamless handoff protocol between agents. Defines the Context Bundle that is passed forward so the developer never re-enters information, and the single handoff banner shown at each transition. Loaded by orchestrator and every phase agent.

• audit-snapshot

  Incremental audit snapshot management for Claude Code. Maintains repository audit snapshots, detects changed and deleted files, supports full versus incremental audits, and tracks token savings by skipping previously audited files.

• background-task-notify

  Notification rule for agents that launch background processes (dev servers, test runners, subagent QA passes). Load in agents that start long-running or async work the developer needs visibility into (orchestrator, setup-agent, qa-agent).

• brownfield-reading

  Anti-speculative reading discipline for existing codebases. MANDATORY for spec-agent brownfield mode and recommended for any agent inspecting code it did not write.

• cache-first-reads

  Token-reduction rule for agents that read project config/specs/knowledge.md repeatedly across a pipeline run. Check context_bundle.md before re-reading raw files from disk. Load in any agent whose tools include Read and that consumes CLAUDE.md, knowledge.md, business_brief.md, or spec files.

• capability-modification

  Defines how Claude Code capabilities evolve safely. Standardizes command modification, orchestrator step injection, agent input/output contracts, runtime context propagation, and persisted state boundaries.

• citation-discipline

  How to source every factual claim in specs, requirements, and ADRs. Use whenever producing any design or requirements artifact.

• coverage-report

  Estimates feature-level test coverage using test discovery and feature-to-test mapping heuristics. Produces coverage classifications and reports for audited repositories.

• diagram-generation

  Guidelines for generating visual layouts, sequencing flows, and diagrams using Mermaid or Excalidraw-JSON formats.

• doc-coverage-map

  Diataxis documentation coverage matrix: for every changed public-surface item, check tutorial/how-to/reference/explanation coverage (✅/❌). Flag zero-coverage gaps and stale architecture diagrams. Used by docs-agent.

• finding-verification-gate

  Audits security findings to compute a threat-confidence-score and suppress false positives.

• gap-analysis

  Skill inventory check run by spec-agent after the developer interview and before spec writing. Derives the task list from interview answers, checks each task against the skill registry in CLAUDE.md §6.1–6.2 and knowledge.md §Project skills, then verifies each matched skill exists on disk. If any tas

• git-operations

  Standardized Git operations for Claude Code agents. Handles branch discovery, checkout workflows, stash management, validation, retries, timeout protection, and conflict surfacing.

• greenfield-grounding

  How to ground greenfield design work — what counts as a source when there is no codebase to cite. MANDATORY for spec-agent greenfield mode and for architecture-agent. The greenfield counterpart of brownfield-reading.

• greenfield-interview

  Routing guide for greenfield interview batches. Points agents to the canonical question bank in setup-interview-questions. Do not store questions here — they live in setup-interview-questions only.

• intake-validation

  Work intake validation criteria and manifest format

• interaction-style

  Shared rules for how agents talk to the developer: batch questions, progress bar format, parallel writes, agent attribution, and Decide Later. Loaded by all pipeline agents.

• interactive-prompts

  Standardized human interaction patterns for Claude Code agents. Provides numbered selection menus, typed input validation, confirmation workflows, retry/cancel behavior, and error prompts.

• knowledge-harvesting

  Extracts and consolidates codebase rules, conventions, and recurring debugging lessons from session histories and commit logs.

• knowledge-md

  How to read, validate, update, and write knowledge.md — the committed per-repo living document that owns all project-specific additions to the SDLC. Covers uninferable facts AND the project registry (agents, skills, routing, tech leads, built features). CLAUDE.md references this file for sections 4.

• live-qa

  Live UI quality assurance: health-score rubric (Console/Links/Visual/Functional/UX/Perf/Content/A11y), WTF-likelihood self-stop, fix-count hard cap, diff-aware/full/quick modes. Chrome DevTools MCP (web) and Maestro MCP (mobile) are the only automation backends — no Playwright, no raw CDP-over-Bash.

• onboarding-friction

  Audits workspace onboarding files, setup guides, and dependency configurations to measure setup friction and Time-To-Hello-World (TTHW) factors.

• parallel-writes

  Speed rule for agents that write, create, or update multiple independent files. Fire all Write/Edit calls in one turn instead of sequentially. Load in any agent whose `tools:` frontmatter includes Write or Edit.

• project-bootstrap

  Stack-agnostic procedure for scaffolding a new project. Used by bootstrap-agent, which pairs it with the stack-specific bootstrap skill (django-bootstrap, node-bootstrap) that owns the exact commands.

• project-permissions

  Configure project tool permissions ONCE during setup instead of prompting repeatedly. Maps a chosen permission profile to .claude/settings.json allow/deny so Read/Write/Edit/Bash for normal pipeline work do not re-prompt. Loaded by setup-agent.

• proposal-evaluation

  MANDATORY procedure whenever any agent detects a missing or inadequate capability (no agent/skill/command covers the work, or an existing one repeatedly falls short). Use this skill instead of ever creating capability or asking the developer "shall I create an agent?".

• retry-policy

  Loop-breaker for agents. After two identical failed attempts at the same action, stop, reassess, and switch strategy or ask the developer. Prevents infinite retry loops. Loaded by every agent that runs commands or edits files.

• root-cause-debugging

  Root-cause debugging discipline: Iron Law "no fix without investigation first", pattern-match table (race/nil/corruption/integration/config/cache), 3-strike hypothesis rule, mandatory regression test. Used by debug-agent via /investigate.

• scope-drift-detection

  Informational scope-drift check: compare spec intent against actual diff file list. Flags CLEAN, DRIFT DETECTED, or REQUIREMENTS MISSING. Never blocks; surfaces for review-agent output.

• security-scans

  Catalogue of security scan types offered at setup and how each maps to a gate command per stack. Lets a project pick multiple scans instead of one. Loaded by setup-agent and security-agent.

• session-sync

  Synchronizes active browser cookies from the developer profile to the headless QA session to bypass SSO, MFA, or auth gates during QA runs.

• setup-interview-questions

  Single source of truth for every question batch in the pipeline. All agents pull from here verbatim — no ad-hoc question generation. Three greenfield batches, one architecture batch, one spec batch.

• spec-paths

  Dynamic spec layout: .claude/specs/<feature>/<change>.md — e.g. todo/add-due-date.md. Used by spec-agent, orchestrator, /approve-spec, /change-feature, /spec-status.

• spec-writing

  Section-by-section quality bar for spec_N.md files. Use alongside the spec_template when drafting any spec.

• specialist-review-panel

  Multi-specialist diff review: Security, Performance, Data-Migration, API-Contract, Testing, Maintainability. Conditional gating by diff signals + line-count thresholds. Dedup by fingerprint, confidence boost on agreement. Feeds into review-agent gate 4.

• threat-modeling

  Attack-surface census, OWASP Top 10 + STRIDE threat model, confidence-scored findings with pre-emit source-quote verification gate and independent re-scorer validation. Used by security-agent in deep-scan mode (/threat-scan).

• ui-brief-alignment

  Design-fidelity rule for any agent that writes or edits UI components, pages, or layout code. Read ux_brief.md before touching UI files, never invent visual elements that conflict with it. Load in frontend developer agents, component builders, page builders — platform or stack layer.

• verification-evidence

  The IRON LAW from gstack: no PASS verdict stands on stale evidence. If code changed after tests ran, re-run before reporting. Also adds diff-coverage audit: confirm every changed file is tested. Used by test-agent.

• version-bump-policy

  Semantic version bump state machine: FRESH/ALREADY_BUMPED/DRIFT_STALE/DRIFT_UNEXPECTED. Diff-size-driven bump class (MICRO auto-patch, PATCH auto, MINOR/MAJOR always ask). Used by release-agent.

• work-req

  Targeted requirement-level build flow. Given only a REQ-ID, looks up the spec in one grep via req-index.md, detects the spec's format, reads only the relevant sections, builds only the delta, and updates all tracking. Handles all three known spec formats. Never reads multiple spec files.

• angular-bootstrap

  Exact scaffold procedure for a new Angular frontend. Used by bootstrap-agent with project-bootstrap skill.

• angular-component

  Angular standalone component implementation checklist.

• angular-form

  Angular Reactive Forms implementation checklist.

• angular-guard

  Angular guard and HTTP interceptor implementation checklist.

• angular-service

  Angular service implementation checklist.

• ngrx-store

  NgRx feature store implementation checklist.

• compose-service-pattern

  Author a production-shaped docker-compose.yml: health-gated dependencies, isolated networks, named volumes, secrets.

• docker-bootstrap

  First-time Docker setup for a project: Dockerfile, .dockerignore, base compose skeleton, build/lint commands.

• image-size-audit

  Scan a built image for CVEs and audit layer/size bloat, recommending minimal-base or layer-consolidation fixes.

• multi-stage-dockerfile

  Write or review a multi-stage Dockerfile with pinned base images, cache-friendly layer order, and a non-root runtime user.

• alembic-safe-migration

  Safe Alembic migration checklist for FastAPI + SQLAlchemy 2.

• fastapi-auth

  FastAPI JWT auth setup — token issuance, verification, refresh, revocation.

• fastapi-bootstrap

  FastAPI project scaffold checklist — deps, layout, async DB, CI.

• fastapi-endpoint

  FastAPI Pydantic schema → service → router → test checklist.

• guard-interceptor-pattern

  Auth guard, RBAC, interceptor, and exception filter checklist for NestJS.

• nestjs-bootstrap

  NestJS project scaffold checklist -- deps, layout, DI wiring, CI.

• new-endpoint

  DTO -> controller -> service -> test checklist for NestJS REST endpoints.

• new-module

  Feature module + service scaffolding checklist for NestJS.

• safe-migration

  Safe TypeORM/Prisma migration checklist for NestJS.

• nextjs-bootstrap

  Greenfield scaffold procedure for a new Next.js 14+ App Router project with Node.js backend.

• nextjs-component

  Procedure for building Next.js App Router pages and React components.

• nextjs-endpoint

  Procedure for creating Next.js Route Handlers and Node.js/Express API endpoints.

• nextjs-server-action

  Procedure for implementing Next.js Server Actions for mutations and form handling.

• bullmq-job

  BullMQ job implementation checklist.

• express-endpoint

  Express/Fastify endpoint implementation checklist.

• jwt-auth

  JWT authentication implementation checklist.

• node-bootstrap

  Node.js project scaffold checklist.

• prisma-safe-migration

  Safe Prisma migration checklist — additive-first, reviewed SQL.

• react-developer

  Activate for React web application development tasks. Trigger phrases: "react component", "react hook", "useState", "useEffect", "react context", "JSX", "react state", "react props", "react router", "tanstack query", "react query", "react memo", "react performance", "react form", "react error bounda

• index-strategy

  Select and safely create the right PostgreSQL index type for an access pattern.

• query-explain-analyze

  Read and act on PostgreSQL EXPLAIN (ANALYZE, BUFFERS) plans to fix slow queries.

• safe-migration-review

  Safe PostgreSQL migration checklist — lock-aware, additive-first, no destructive ops without approval.

• schema-normalization-check

  Checklist for PostgreSQL type/constraint/normalization choices before DDL is finalized.

• python-async

  Python asyncio implementation checklist — async/await, concurrency limits, resource cleanup.

• python-cli

  Python CLI implementation checklist — Click/Typer, options, exit codes, tests.

• python-pipeline

  Python data pipeline implementation checklist — Pydantic, chunking, idempotency.

• python-script

  Python script implementation checklist — pathlib, logging, typed, idempotent.

• react-api-client

  Procedure for building a typed Axios API client layer in React.

• react-auth
• react-bootstrap
• react-component
• react-state
• pinia-store-pattern
• vue-bootstrap
• vue-component
• vue-composable
• vue-router

Hooks (3)

• PreToolUse

  d="${CLAUDE_PROJECT_DIR:-.}"; [ -f "$d/.claude/memory/orchestrator_state.md" ] || grep -qs 'approvals/ACTIVE' "$d/.claude/settings.json" || exit 0; f=$(sed -n 's/.*"file_path"[[:space:]]*:[[:space:]]*

• PreToolUse

  d="${CLAUDE_PROJECT_DIR:-.}"; [ -f "$d/.claude/memory/orchestrator_state.md" ] || grep -qs 'approvals/ACTIVE' "$d/.claude/settings.json" || exit 0; [ -f "$d/.claude/gate-off" ] && exit 0 f=$(sed -n 's

• Stop

  d="${CLAUDE_PROJECT_DIR:-.}/.claude/memory"; s="$d/orchestrator_state.md"; g="$d/gate_results.md"; if [ -f "$s" ]; then echo 'Reminder: verify orchestrator_state.md reflects the last completed step.';

Similar plugins