divyamohan1993/dmjcustomizations
divyamohan1993/dmjcustomizations · 0 stars · 1 plugin
Marketplace
Install
/plugin marketplace add divyamohan1993/dmjcustomizationsRun inside Claude Code, then install plugins from it. Check the repo's README before you run it.
Plugins 1
After adding the marketplace, install one with /plugin install <name>@dmj.
- 1dmjDivya's parallel-first agentic engineering skills: teammate fan-out workflows, adversarial verification, security and performance gates from line 1
/plugin install dmj@dmj
Files
dmjcustomizations
Research-grade, production-deployable agentic software engineering framework for Claude Code and Codex. Skills = policy. Hooks = enforcement. Generated gates = verification. Teams = execution. Terse, date-agnostic, adversarially verified, secure from line 1.
Why
Built for the current Claude generation: teammate fan-out, cheap code, models trusted with judgement, hard gates where judgement is not enough. Adopted frameworks bind everything AI writes into a target repo, actively: EARS requirements, ASD-STE100 prose, Gherkin acceptance, OWASP ASVS L2, WCAG 2.2 AAA, Keep a Changelog, conventional commits, DPDP + GDPR.
- Parallel between gates, serial at gates. Teammates fan out; the user approves at a few hard checkpoints. Nothing else blocks.
- Evidence over claims. Fresh-context teammates adversarially verify specs, code, and every "done". Self-review is not review.
- Demonstrate, don't describe. Competing approaches become disposable worktree spikes with benchmarks, not paragraphs.
- Dynamic by design. No hardcoded models or dates; the rare version number a reference file carries comes with a re-check-at-invocation instruction. Skills probe for the strongest model and newest stable tooling at invocation time.
- Security and performance from line 1. Threat models at design time, quantum-safe crypto defaults, O(1)-first thinking, budgets the skills require projects to enforce in CI.
- Terse. Every skill has a context budget. Each rule lives once at its owning home. A runaway cap is only a backstop. The always-loaded meta-skill has the tightest cap. Substance is never cut just to hit a number.
Skills
| Skill | One-liner |
|---|---|
| using-dmj | Meta-skill, injected each session: how skills are found, routed, and prioritized |
| orchestrating-products | The wireframe: idea to shipped ecosystem through staged gates, each stage on the cheapest tier that holds its bar |
| brainstorming | Idea to approved spec: parallel context sweep, batched questions, evidence-based options |
| writing-plans | Spec to implementation plan with per-task dependencies, parallelizability, acceptance criteria |
| executing-plans | Team executes a plan: tasks claimed concurrently, worktree isolation, review gates |
| team-driven-development | Plan execution in the current session with implementer and reviewer teammates |
| dispatching-parallel-teams | Any 2+ independent tasks: fan out a team, coordinate, synthesize |
| test-driven-development | Iron Law TDD plus an extreme edge-case taxonomy (adversarial, concurrency, boundaries) |
| systematic-debugging | Root cause before fixes; parallel hypothesis investigation |
| verification-before-completion | Evidence before claims; independent fresh-context verification |
| requesting-code-review | Multi-lens parallel review panel: correctness, security, performance, simplicity |
| receiving-code-review | Rigor over performative agreement; verify before implementing feedback |
| using-git-worktrees | Isolation for parallel work and disposable spikes |
| finishing-a-development-branch | Verified finish: merge, PR, or cleanup, with team and worktree teardown |
| shipping-to-production | Scripted, reversible deploys: idempotent script, health-probe gate, one-step rollback, priced resources |
| equipping-projects | Detect-and-wire project guard rails in minutes: hooks, CI, MCP, preview, derived from the repo's signals |
| stewarding-data | Data outlives code: drilled restores, expand-migrate-contract, tested down-paths, retention and deletion by design |
| observing-production | Health from signals, not silence: SLOs, symptom alerts, correlation IDs, blameless post-mortems that land as commits |
| writing-skills | TDD for documentation: baseline, write, close loopholes, team-tested |
| evolving-skills | Gated loop to improve skills from confirmed learnings: TDD proposal, gates, human-merged PR, never autonomous |
| defending-in-depth | Security from line 1: threat model, OWASP, zero trust, quantum-safe defaults |
| enforcing-performance-budgets | O(1)-first, measured budgets, cost as a budget, cache-first, regressions block merges |
| researching-deeply | Parallel research with adversarial source verification and dated citations |
| exploring-codebases | Five-lens parallel codebase mapping with an anti-redundancy gate: reuse before rebuild |
| tracing-codebases | Parallel slice-by-slice tracing of how a codebase really works, explained in chat, no artifacts |
| harnessing-claude | Capability routing: strongest native Claude feature for every job, never hand-rolled substitutes |
| crafting-experiences | Experience supremacy: Jobs test, first-second hook, cinematic with purpose, zero user burden, web-URL delivery default |
| art-directing | Research-driven visual identity: psychology-led palette, distinctive type, edge-to-edge responsive, screenshot-verified |
| landing-sessions | Session teardown: learnings to memory, state to git, resources to zero, threads surfaced |
| selling-the-vision | Launch persuasion: one-second hook, keynote arc, one message, one CTA, honest claims |
| enforcing-quality-gates | Every repo carries a generated gate: unit, acceptance, mutation, coverage, complexity, fuzz, and security checks. Slow lanes run later, not disabled. |
| humanizing-output | Pre-push gate every project carries: strips em dashes and AI-tell language, reviewed humanize, prose that never reads AI-generated |
Install
The plugin installs as dmj. Claude skills fire as dmj:brainstorming, dmj:test-driven-development, etc. Codex loads the native .codex-plugin/plugin.json and .codex/hooks.json surfaces; it does not use the Claude manifest as a substitute.
Claude Code
From GitHub:
/plugin marketplace add divyamohan1993/dmjcustomizations
/plugin install dmj@dmj
Codex
From the Git marketplace:
codex plugin marketplace add divyamohan1993/dmjcustomizations codex plugin add dmj@dmj
The Codex install commands update Codex-managed local state. Repository validation and hooks do not edit local config or cache.
Or from a local clone:
/plugin marketplace add D:\dmjcustomizations
/plugin install dmj@dmj
Surfaces
| Surface | How it loads |
|---|---|
| Claude Code CLI, desktop, VS Code | The install above; one config serves all three |
| Codex CLI | The Codex install above; native manifest and hook routes are checked in |
| Claude Code on the web | Automatic: this repo commits .claude/settings.json declaring the marketplace and plugin, installed at session start |
| Cowork | Customize, Plugins, Add marketplace, divyamohan1993/dmjcustomizations (Cowork reads your claude.ai account, not ~/.claude) |
| Claude.ai chat | No plugins on that surface. Run node scripts/export-claude-ai.mjs, upload the zips you need under Settings, Capabilities |
| Agent SDK | plugins: [{ type: "local", path: "<clone>" }] |
The git-safety guard uses Perl when present and falls back to Node. If neither tool is available, the guard denies the action and reports the missing dependency.
Provenance
Forked from obra/superpowers 5.1.0 by Jesse Vincent (MIT). Rebuilt 2026-06-10 by Divya Mohan. MIT licensed.
{
"name": "dmj",
"description": "Personal marketplace for the dmjcustomizations skills library",
"owner": {
"name": "Divya Mohan",
"email": "divyamohan1993@gmail.com"
},
"plugins": [
{
"name": "dmj",
"description": "Divya's parallel-first agentic engineering skills: teammate fan-out workflows, adversarial verification, security and performance gates from line 1",
"version": "5.1.0",
"source": "./",
"relevance": {
"topic": "parallel-first engineering discipline",
"signals": {
"cli": [
"git",
"pnpm",
"npm",
"cargo",
"go"
],
"filesRead": [
"**/CLAUDE.md",
"**/package.json",
"**/Cargo.toml",
"**/go.mod"
]
}
},
"author": {
"name": "Divya Mohan",
"email": "divyamohan1993@gmail.com"
}
}
]
}Facts
- Kind
- Marketplace
- Repo
- divyamohan1993/dmjcustomizations
- Group
- Uncategorized
- Stars
- 0
- Marketplace name
- dmj
- Owner
- Divya Mohan
- Language
- Shell
- Created
- 2026-06-10
- Last push
- 2026-08-23
- Plugins
- 1
- 1obra/superpowersobra/superpowersAn agentic skills framework & software development methodology that works.297.9k
- 2affaan-m/everything-claude-codeaffaan-m/ECCThe agent harness performance optimization system. Skills, instincts, memory, security, and research-first development for Claude Code, Codex, Opencode, Cursor and beyond.277.5k
- 3anthropics/skillsanthropics/skillsPublic repository for Agent Skills180.6k
- 4f/prompts.chatf/prompts.chatf.k.a. Awesome ChatGPT Prompts. Share, discover, and collect prompts from the community. Free and open source — self-host for your organization with complete privacy.172.7k
- 5anthropics/claude-codeanthropics/claude-codeClaude Code is an agentic coding tool that lives in your terminal, understands your codebase, and helps you code faster by executing routine tasks, explaining complex code, and handling git workflows - all through natural language commands.150.3k
- 6nextlevelbuilder/ui-ux-pro-max-skillnextlevelbuilder/ui-ux-pro-max-skillAn AI skill that provides design intelligence for building professional UI/UX across multiple platforms.135k