architecture

Architecture rules for NEMESIS, a software project divided into separate layers for the interface, application logic, AI agents, data storage, and external tools. The rules define which layers may depend on one another and how agents communicate.

In plain words
What is it for?
Use it to guide implementation of NEMESIS features, agent classes, typed events, project state, database access, and command-line tool wrappers.
Why use it?
Strict boundaries make it clearer where code belongs and help prevent parts of the application from becoming tightly coupled. They also give agents a consistent way to pass results and events between components.

Cursor rule for Cursor

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.

agentmods
npx agentmods add rules/joaovicdev/nemesis/architecture
Clone the repo
git clone --depth 1 https://github.com/joaovicdev/nemesis

Made for: Cursor.

Per session 556 This file is loaded in full into every session.
When invoked 556 The same file — it is already loaded in full.
Security scan A 0 findings. Scan, not verified.
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 $0.00556 $0.00556
Opus 5 $0.00278 $0.00278
Sonnet 5 $0.00111 $0.00111
Haiku 4.5 $0.00056 $0.00056

Measured yesterday against content hash 33a7f9dce447, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

architecture 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 yesterday.

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/architecture.mdc · 71 lines

How it starts

The opening of the file, as written. The whole thing — 71 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Architecture Rules

Layer Separation

NEMESIS is divided into four strict layers. Imports must only flow downward — upper layers depend on lower ones, never the reverse.

tui/          ← presentation only, no business logic
core/         ← domain models, config, project state
agents/       ← orchestrator, analyst, executor (AI + tool logic)
db/ + tools/  ← persistence and external tool wrappers
  • tui/ MUST NOT import from agents/ directly. All agent interaction goes through core/.
  • agents/ MUST NOT import from tui/. Agents communicate results via events or callbacks registered by core/.
  • tools/ contains only thin wrappers that execute system commands and return raw output. No analysis or interpretation belongs here.

Agent Patterns

  • All agents (orchestrator, analyst, executor) are async classes.
  • Agents communicate results by emitting typed Finding or AgentEvent objects.
  • Executors are short-lived: instantiated per task, execute, emit result, and are garbage-collected.
  • The Orchestrator holds the only long-lived context object (ProjectContext).
  • The Analyst ALWAYS processes raw tool output before it reaches the Orchestrator. Raw subprocess output must never be sent directly to the LLM as-is.

Data Flow

User input → Orchestrator → [spawns] Executor(s)
                                         ↓ raw output
                                      Analyst
                                         ↓ structured Finding
                         Orchestrator ← Finding
                                ↓
                          ProjectContext (SQLite)
                                ↓
                            TUI update

Destructive Action Gate

Any tool or agent action that is potentially destructive (exploit execution, brute force, file upload, lateral movement) MUST:

  1. Be tagged with destructive=True in its tool definition.
  2. Emit a ConfirmationRequired event to the TUI before execution.
  3. Only proceed after receiving an explicit UserConfirmed event.
  4. Log the confirmation with timestamp to the project session.

Read the full file on GitHub · 71 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. yesterday First seen · 71 lines · 556 tokens per session scan A 33a7f9dce447

Subscribe to this mod's changes

architecture is a cursor rule published in the GitHub repository joaovicdev/nemesis (4 stars, last pushed 4mo ago), licensed MIT. It adds 556 tokens to every session, about $0.0028 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 2026-08-31.