subagents

subagents is a cursor rule for Claude Code, Cursor from rajitsaha/100xprism. It costs 26 tokens per session (1,350 once invoked), scanned A, original, MIT.

Guidance for using subagents, which are separate helpers that handle parts of a larger coding task.

In plain words
What is it for?
Use it for codebase exploration, research, large refactors, parallel analysis, and test writing.
Why use it?
It helps divide complex or parallel work while keeping the main coding conversation focused.

Cursor rule for Claude CodeCursor

Written for Claude Code and Cursor: PreToolUse hook event, but also installed under .cursor/. Also seen: mentions subagents; mentions Claude Code.

Good fit Use it for codebase exploration, research, large refactors, parallel analysis, and test writing.

Compare 6 cursor rules from other repositories ↓
Install with agentmods
npx agentmods add rules/rajitsaha/100xprism/subagents
Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

Clone the repo
git clone --depth 1 https://github.com/rajitsaha/100xprism

Made for: Claude Code, Cursor.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

agentmods badge for subagents

README.md
[![agentmods](https://agentmods.dev/badge/rules/rajitsaha/100xprism/subagents/github.svg)](https://agentmods.dev/rules/rajitsaha/100xprism/subagents)
Your own site
<a href="https://agentmods.dev/rules/rajitsaha/100xprism/subagents"><img src="https://agentmods.dev/badge/rules/rajitsaha/100xprism/subagents/github.svg" alt="Measured on agentmods" height="20"></a>

Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.

agentmods 80×15 button for subagents

Your own site · 80×15
<a href="https://agentmods.dev/rules/rajitsaha/100xprism/subagents"><img src="https://agentmods.dev/badge/rules/rajitsaha/100xprism/subagents.svg" alt="Reviewed on agentmods" width="80" height="20"></a>
Per session 26 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 1,350 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5.1 $0.00026 $0.01350
Opus 5 $0.00013 $0.00675
Sonnet 5 $0.00005 $0.00270
Haiku 4.5 $0.00003 $0.00135

Measured 5d ago against content hash c048671e7617, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-08, from the pricing page.

Security

Grade A, and why

subagents scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 5d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

.cursor/rules/subagents.mdc · 123 lines

How it starts

The opening of the file, as written. The whole thing — 123 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Invoke this skill when a task is complex, exploratory, or parallel in nature. Subagents keep the main context clean and focused.

When to Use Subagents

  • Append "use subagents" to any request where you want Claude to throw more compute at the problem
  • Use for: codebase exploration, parallel analysis, research, large refactors, test writing
  • Do NOT use for: simple one-file edits, quick lookups, single-step tasks

Three Subagent Strategies

A. Parallelization

Offload independent tasks to separate subagents running simultaneously:

use 5 subagents to explore the codebase
→ Explore entry points and startup
→ Explore React component structure
→ Explore tools implementation
→ Explore state management
→ Explore testing infrastructure

B. Context Isolation

Offload heavy research or analysis to a subagent so the main agent's context stays focused on the implementation.

  • Research → subagent → returns a structured result (see Standard return contract below)
  • Main agent uses the result, never sees the raw noise

Reframe for the 1M-context era: with a 1M-token window, token savings is no longer the main reason to isolate. Isolate for parallelism (independent work running at once) and adversarial independence (a fresh subagent with zero authoring context reviews work without self-review bias). Treat context savings as a side benefit, not the goal.

C. Permission Routing via Hook

Auto-approve obviously-safe tool calls so they don't interrupt with a prompt, while anything risky still falls through to human review. 100xprism ships this as a real, installable artifact — not just advice:

  • Artifact: ~/100xprism/hooks/permission-router.py (a PreToolUse Bash hook).
  • Tier 1 (offline): a deterministic allowlist auto-approves read-only commands (ls, cat, git status, grep, …); destructive/network/credential commands are never auto-approved.
  • Tier 2 (optional): set HOOK_ROUTER_MODEL=claude-haiku-4-5 (needs the claude CLI) to route ambiguous commands to a cheap model; only a confident "safe" verdict grants permission, so escalation to a deeper model (e.g. Opus 5) or a human is the default for anything uncertain. The router never blocks — it only grants.
  • Enable it: re-run the installer and turn on the permission-router hook (it ships off by default), or run python3 ~/100xprism/adapters/lib/modules.py emit-hooks with HOOK_ROUTER=1.
  • See ~/100xprism/hooks/README.md and the hooks docs: https://docs.claude.com/en/docs/claude-code/hooks

Read the full file on GitHub · 123 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 5d ago First seen · 123 lines · 26 tokens per session scan A c048671e7617

Subscribe to this mod's changes

subagents is a cursor rule published in the GitHub repository rajitsaha/100xprism (10 stars, last pushed 8d ago), licensed MIT. It adds 26 tokens to every session and 1,350 once invoked, about $0.0001 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-09-03.