107 ready-made Claude agents — code review, copywriting, analytics, support, growth, legal. Drop them into Claude Code and watch the right one auto-route to your task in under 5 seconds.
⚡ Get Instant Access for $19 →
These aren't mockups. Each example below is the actual format you get back when you trigger the named subagent in Claude Code. The agent reads the description, decides it's the right one, and produces output in a structured shape you can copy-paste into a PR, ticket, or doc.
🔴 Blocking — fix before merge 1. app/api/auth.ts:42 — JWT secret hardcoded in source. Move to env. 2. app/api/upload.ts:15 — SSRF: fetch URL is user-controlled, no allowlist. 3. lib/db.ts:88 — SQL built via string concat. Parameterize. 🟡 Should fix 1. middleware.ts:23 — rate limit only by IP, trivially bypassed by VPN. 2. app/api/verify/route.ts:67 — error message leaks DB schema.
## Hypotheses (likeliest first) 1. Weekly cron clobbers DB connections — test: align cron schedule against the 500 timestamps. Look for a job that fires Fri 5:55pm. 2. Friday traffic spike exhausts pool — test: pool metrics + latency p99 around the failure window. 3. Time-zone bug treating Fri evening as off-hours — test: grep the handler for any Date comparisons or business-hours guards. ## Next observation Pull the last 4 Fridays' 500s with timestamps + correlation IDs. If they cluster in a 5-min window → it's the cron. Spread across hours → pool. Mixed with timezone strings → the TZ bug.
Hook 1 — price flex Spent ₹40k on ChatGPT Plus this year? 107 ready-made Claude agents for ₹499. One-time. Lifetime. Hook 2 — workflow Drop one file. Claude becomes your QA, copywriter, analyst, support team. 107 agents. 5-second install. Hook 3 — contrarian Stop writing prompts. Drop subagent files instead. Claude auto-routes the right expert in 5 seconds. ₹499 once.
Engineering, content, marketing, business, data, productivity, legal, India-context, personal — 9 categories of opinionated, ready-to-fire agents.
5-minute install guide plus a deep reference manual covering customization, chaining, troubleshooting, and FAQ.
Working TypeScript project (customer-research-agent) built on the Anthropic SDK. Run it, fork it, ship it.
v1.1 ships with 30+ more agents (sales, customer success, design, recruiting) plus 2 more SDK starters — free for existing buyers.
A small sample of the 107:
Every agent in the pack is a single markdown file. YAML frontmatter tells Claude when to use it and what tools it can touch; the body is the system prompt. Plain text. Read it, edit it, version-control it.
--- name: code-reviewer description: Use when reviewing a diff, pull request, or commit before merge. Focuses on correctness and footguns, not nit-pickery. tools: Read, Grep, Glob, Bash --- You are a senior code reviewer. Your job is to catch the things a careless review would miss, while ignoring things that don't actually matter. ## What to look for, in order 1. Correctness. Does the code do what the diff claims? Look at edge cases the author probably skipped: empty inputs, nulls, concurrent access, error paths, off-by-one, time zones. 2. Security. Anything that touches user input, auth, secrets, or external services: assume it's wrong until proven otherwise. 3. Concurrency & state. Race conditions, missing locks, stale caches, non-idempotent operations called from retried paths. ... (continues — full file is ~120 lines)
No accounts to create. No API keys. No new tools to learn. If you already have Claude Code installed, you're 60 seconds away from your first agent firing.
User-scoped install (agents available across every project). The full breakdown is below if you want it.
unzip agents-bundle.zip && cp -r agents-bundle/subagents/* ~/.claude/agents/
Agents live inside one repo, version-controlled with the code. Anyone who clones the repo gets them.
cd /path/to/your/repo mkdir -p .claude/agents cp -r /path/to/agents-bundle/subagents/* .claude/agents/
Claude auto-routes by reading every agent's description field. You can also call one explicitly.
# Auto-route (most common) > "Review the last commit for security issues." → security-auditor # Explicit > "Use bug-hunter to figure out the 500 on /api/x."
Each agent is plain markdown. Open it, sharpen the system prompt with your codebase / brand voice / customer context. The agents get noticeably better the more they know about your specific work.
code ~/.claude/agents/code-reviewer.md # edit body, save, done — Claude picks it up on next message
| User-scoped | Project-scoped | |
|---|---|---|
| Where files live | ~/.claude/agents/ | <repo>/.claude/agents/ |
| Works across projects | Yes | No — this repo only |
| Shared with collaborators | No (just you) | Yes (via git) |
| When to use | Solo work, all projects | Team workflows |
You can use both. Project-scoped agents with the same name override user-scoped ones — handy when you need a tuned variant just for one codebase.
Claude reads your full instruction and orchestrates the hand-off between agents. Output from one becomes context for the next. Three real chains people run with this pack:
“Use bug-hunter to find the root cause of the 500 on /api/x. Once you've isolated it and I've written the fix, use test-writer to lock in a regression test, then code-reviewer on the diff before I merge.”
“Use competitor-analyst on Stripe, Razorpay, and PayU to find the positioning gap. Pass that brief to ad-copywriter for 3 Meta variants and landing-page-copywriter for the above-fold. Then brand-voice-coach reviews everything for tone consistency.”
“Use market-sizing-analyst to give me an honest TAM/SAM/SOM for tier-2 Indian cities. Then customer-research-interviewer designs the discovery interview script. After I run 8 interviews, gtm-strategist proposes the launch motion based on what we learned.”
Each category is curated — not a dump of every possible agent, but the ones that earn their keep. Each agent is opinionated, focused on one job, and ready to fire.
Engineering
15 agents
Content
12 agents
Marketing
11 agents
Business
13 agents
Data
12 agents
Productivity
14 agents
Legal
10 agents
India Context
5 agents
Personal
14 agents
Subagent files are markdown + YAML. They work natively in Claude Code; the prompt content works anywhere else you can paste a system prompt. The bundle's SDK starter is a separate, self-contained TypeScript project for the autonomous-agent case.
Native fit. Drop into ~/.claude/agents/ — Claude auto-routes by description. The bundle's primary target.
Copy the body of any agent file (everything below the frontmatter) into a Project's custom instructions. Auto-routing isn't available, but the prompt content works identically.
Same flow as Claude Code — agents go into the project's .claude/agents/ folder and the whole team gets them. Subagent routing works out of the box.
The bundle includes a working TypeScript starter (customer-research-agent) built on the SDK. Fork it for autonomous agents that run on schedules, webhooks, or as part of a larger app.
All 2,000+ skills, 107 agents, 50 pre-built Claude.ai Projects, every guide & the SDK starter. One purchase, lifetime access.
107 agents covering every role you wish you could afford. $19 once. Made for Claude Opus 4.8 — Mythos Level Intelligence inside.
⚡ Get Instant Access for $19 →One-time payment · Lifetime access · Instant delivery