100xprism

100xPrism

Stop vibe coding. Ship production-grade software.

One source of truth. 67 modules generate native config for Claude Code · Cursor · Codex. Quality gates run on every commit.

Install

npm (any platform — macOS, Linux, Windows):

npm install -g 100xprism && 100xprism install

curl (macOS / Linux):

curl -fsSL https://raw.githubusercontent.com/rajitsaha/100xprism/main/get.sh | bash

Either way, 100xprism install clones the toolkit to ~/100xprism and provisions your AI tools. The npm package is a thin launcher — the modules, hooks, and plugins live in that clone, which 100xprism update keeps current. Before installing, 100xPrism cleans legacy startup hooks, stale command links, and any old owned dashboard process. Install does not edit ~/.zshrc, ~/.bashrc, or ~/.bash_profile. If you want optional aliases for the current terminal only, run source ~/100xprism/shell/aliases.sh.

Windows: plugin sync works, but native Windows module emit is being reworked (#54). For full module support today, install under WSL with either method above.

Set up a project:

cd your-project && 100xprism init

Keep up to date:

100xprism update                    # pull latest, then add/update/remove skills + plugins
100xprism update --plugins-only     # refresh plugins only (repo already current)
100xprism uninstall                 # stop dashboard + remove legacy shell-startup entries/symlinks
npm install -g 100xprism@latest     # (optional) upgrade the launcher itself

install and update are fully reconciling, not append-only — every run:

• adds newly shipped skills, slash commands, and curated plugins,
• updates changed ones in place, and
• removes skills, slash-command aliases, and 100xprism-managed plugins that were deleted or merged upstream.

Your own hand-authored skills/commands and any plugins you enabled yourself are never touched. See docs/USAGE.md for details.

Cloned to a custom path? The default install lives at ~/100xprism. If you cloned elsewhere, update your shell + Claude Code config — see Custom install location.

The pipeline

/understand → /context → /issue → /spec → /fix → /commit
                                                    ↓
              /techdebt ← /gate → /grill → /pr → /push → /release

Every /commit and /push runs a 5-point gate — tests, security, build, Docker, cloud. Nothing ships without passing.

What you get

67 modules	27 slash commands + 40 auto-trigger skills — see full reference below
14 Claude Code plugins	superpowers, playwright, github, hookify, claude-mem, understand-anything, ui-ux-pro-max, motion-framer, and more
7 database engines	Postgres, Cloud SQL, Snowflake, Databricks, Athena, Presto, Oracle — one /db interface
27 SaaS CLIs	/connect installs + authenticates GitHub, AWS, Stripe, Supabase, and more from .env
4 project templates	node-fullstack · node-frontend · python-api · docker-compose
CI/Release pipelines	Drop-in GitHub Actions for lint + real-DB tests + E2E + semantic-release

Token, delivery & value economics

For a while, the goal was singular: make spec-driven, agentic development as autonomous as possible — let the agent plan, build, gate, and ship with less and less human steering.

That part is mostly working. The harder, more important question is the one that comes next: what is all this autonomy costing, what did it help deliver, and what was that delivery worth? Every agent run spends real money in tokens. The dashboard measures cost and observable engineering delivery separately; business or human value remains explicitly unmeasured until an outcome source is connected.

This is a first, deliberately humble attempt to make the measurement chain visible to everyone, because watching it isn't one person's job. It's everybody's responsibility.

100xprism tokens   # token economics — what every session, across every repo, is costing
100xprism value    # delivery economics — observable work associated with that spend

• 100xprism tokens — one offline, machine-wide dashboard at a single URL: the input/output/cache split, a startup-bloat meter, an estimated code-vs-files-read-vs-logs-vs-chat composition, and $ cost — by project, model, session, and skill. Claude Code and Codex provide exact local token counters; Cursor agent-transcript JSONL and Antigravity local artifacts contribute project/session/activity coverage but are never assigned invented token cost because their local formats expose no counters. Cursor chats, state.vscdb, and legacy transcript .txt are outside the collector's scope. It auto-refreshes every 30 seconds once started. Start it explicitly with 100xprism tokens, 100xprism dashboard, or 100xprism install --dashboard; shell startup never starts it.
• 100x-value — the dashboard joins exact local token counters to observable delivery evidence. It shows every directory that consumed tokens plus agentic projects discovered machine-wide via marker files. Git supplies commits, deduplicated merged PRs, releases, files, insertions, and deletions; non-repos use an explicitly labeled filesystem-mtime estimate. Directories from unsupported tools show — cost, never a misleading $0. Delivery unit costs use attributed spend only, show attribution coverage, and are never presented as business ROI.
• Budgets, provenance, recommendations & GitHub PR insights — optional daily/weekly limits drive dashboard and shell alerts. A provenance strip shows pricing coverage, outcome-join coverage, source counts, and date range. Recommendations are ranked by estimated opportunity and pair the observed evidence with an action that preserves AI-native autonomy. Optional GitHub CLI integration can fetch PR metadata for locally checked-out remotes plus configured users/repos.

Full guide: docs/token-optimization.md.

Slash commands

The following 27 slash commands are available. Run them inside Claude Code. In Codex, use the generated repo skill by name instead, for example $gate, $commit, or /skills.

Lifecycle

Command	What it does
/branch	Create a conventional feature branch (feat/, fix/, chore/)
/commit	Gate → stage → conventional commit
/grill	Adversarial code review before opening a PR
/pr	Gate → push branch → create PR
/push	Gate → push → monitor CI → verify production health
/release patch|minor|major	Semantic versioning + publish to PyPI/npm/Docker Hub
/launch	Full deploy pipeline in one command

Quality

Command	What it does
/gate	Mandatory 5-point quality gate (tests, security, build, Docker, cloud)
/test	All test layers (unit, integration, E2E) — loops until 95% coverage
/lint	Auto-detect and fix all lint errors (ESLint, TypeScript, ruff)
/security	Vulnerability + secrets scan, auto-fix where possible
/cloud-security	GCP IAM, networking, PII, and compliance scan
/eval	Run module evals — check triggers and output quality

Engineering

Command	What it does
/spec	Turn a vague request into an implementation-ready spec
/fix	Autonomous bug fixer — CI failures, docker logs, Slack pastes
/orchestrate	Plan-first methodology for complex multi-step tasks
/techdebt	Dead code, duplication, stale TODOs
/context	7-day git + GitHub activity dump — orient before coding
/update-claude	Write a CLAUDE.md rule after any correction

Data & Infrastructure

Command	What it does
/db	Query any of 7 database engines from one interface
/query	Plain-English analytics — describe what you want, get SQL
/connect	Install + auth 27 SaaS CLIs from .env

Documentation & Architecture

Command	What it does
/docs	Detect code changes and update documentation
/issue	Investigate a bug and create a detailed GitHub issue
/architect	Architectural Q&A and decision matrices
/enterprise-design	Full technical blueprint — IA, API, data model, stack

Auto-trigger skills (39)

These modules activate automatically when you describe a relevant task — no slash command needed.

Category	Modules
Marketing copy	copywriting, copy-editing, cold-email, email-sequence, ad-creative, social-content
SEO	seo-audit, ai-seo, programmatic-seo, schema-markup, site-architecture
CRO & conversion	page-cro, signup-flow-cro, onboarding-cro, form-cro, popup-cro, paywall-upgrade-cro
Growth & strategy	content-strategy, marketing-ideas, marketing-psychology, launch-strategy, referral-program, churn-prevention, free-tool-strategy, ab-test-setup, analytics-tracking, pricing-strategy
Sales	sales-enablement, competitor-alternatives, paid-ads, revops, product-marketing-context
Design	enterprise-design, visual-system-architect, interaction-engineer, figma-translator
Engineering	subagents, terminal-setup

How it works in your tool

Tool	Generated artifact	Auto-trigger?
Claude Code	~/.claude/skills/<slug>/ + slash command aliases	Yes — per description
Cursor	.cursor/rules/<slug>.mdc (one file per module)	Yes — per description
Codex	AGENTS.md + .agents/skills/<slug>/ + .codex/hooks.json	Yes — repo skills

Every supported tool loads module bodies on demand rather than inlining them, so the always-on context stays small. tier: core marks the modules Cursor keeps resident (alwaysApply: true); everything else is fetched when its description matches. Claude Code plugins remain Claude-specific; use Codex /plugins for Codex-native plugins.

Removed in v3.0.0: Windsurf, Copilot, Gemini, and Antigravity. Those adapters emitted a single concatenated file (~60K chars) that sat in context on every turn — the opposite of progressive disclosure.

This deletes files in your projects. 100xprism update removes the .windsurfrules, GEMINI.md, ANTIGRAVITY.md, and .github/copilot-instructions.md it previously generated from every project in ~/.100xprism/tracked-projects; 100xprism init does the same for the project it runs in.

Only files carrying the Generated by 100xprism header are touched, so a hand-written file of the same name is left alone. If you edited a generated file, it is still removed — but every removal is copied to ~/.100xprism/removed-artifacts/<timestamp>/ first, and is only removed once that copy succeeds. These files are also normally committed, so deletions show up in git status for you to review.

Common CI traps it fixes

npm install 404 inside Docker · useState(false) opacity-0 breaking Playwright · integration tests silently excluded from the gate. Full breakdown →

More

• Full usage guide — daily patterns, multi-project setup, CI templates, project config, troubleshooting
• Architecture — why modules replaced workflows + skills
• Token usage & optimization — audit your plugin/skill footprint and monitor token spend with the local dashboard
• Changelog · Roadmap · Issues

Built by Rajit Saha · 23 years in enterprise data at Udemy, Experian, LendingClub, VMware, Yahoo

If this saves you time, star the repo.