security

A security rule set for NEMESIS, a security-analysis project. It covers private data, which targets may be tested, and how output from tools such as nmap and nikto must be handled.

In plain words
What is it for?
Use it when running security tools, validating IP addresses and domains, storing findings locally, or deciding whether external language-model services may receive data.
Why use it?
It reduces the risk of exposing credentials or scan details, testing systems outside the approved scope, or treating untrusted tool output as safe.

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/security
Clone the repo
git clone --depth 1 https://github.com/joaovicdev/nemesis

Made for: Cursor.

Per session 512 This file is loaded in full into every session.
When invoked 512 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 $0.00512 $0.00512
Opus 5 $0.00256 $0.00256
Sonnet 5 $0.00102 $0.00102
Haiku 4.5 $0.00051 $0.00051

Measured yesterday against content hash e513c64c7d85, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

security 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/security.mdc · 57 lines

How it starts

The opening of the file, as written. The whole thing — 57 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Security Rules

Sensitive Data Handling

  • NEVER log target IPs, hostnames, credentials, CVE payloads, or finding details at DEBUG level or above in plain text.
  • NEVER include raw tool output in exception messages or tracebacks.
  • All findings stored in SQLite are local-only — never transmitted to external services unless the user has explicitly configured and consented to an external LLM provider.
  • Ollama runs locally — this is the default and preferred mode. External provider support (when added) must be opt-in with explicit acknowledgment from the user.

Scope Validation

  • Before any tool execution, the target MUST be validated against the project scope:
    • IP addresses must fall within the configured CIDR ranges.
    • Domains must match or be subdomains of the configured targets.
  • Out-of-scope targets MUST be rejected with a clear error — never silently skipped.
  • Scope validation lives in core/project.py and is the single source of truth.

Tool Output Trust

  • All output from external tools (nmap, nikto, gobuster, etc.) is UNTRUSTED until processed by the Analyst agent.
  • Never interpolate raw tool output directly into shell commands (no shell injection).
  • Always use subprocess with shell=False and explicit argument lists.
  • Sanitize any user-provided input before it is passed to subprocess arguments.

Destructive Action Guardrails

  • Tools tagged destructive=True require explicit user confirmation (see architecture rules).
  • The following actions are ALWAYS considered destructive regardless of tool tag:
    • Any exploit execution
    • Brute-force / credential stuffing
    • File uploads to target systems
    • Commands that modify target system state
    • Any action that could cause service unavailability (DoS-like behavior)
  • Destructive action confirmations MUST be logged with:
    • Timestamp
    • Action description
    • User confirmation ("yes" response)
    • Session ID

Ethical Guardrails

  • NEMESIS is designed for authorized penetration testing only.
  • The project scope acts as the technical enforcement of authorization boundaries.
  • Never add features that bypass scope validation or skip confirmation gates.
  • The disclaimer in README.md and the first-run message are not optional — they must always be present and visible.

Read the full file on GitHub · 57 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 · 57 lines · 512 tokens per session scan A e513c64c7d85

Subscribe to this mod's changes

security is a cursor rule published in the GitHub repository joaovicdev/nemesis (4 stars, last pushed 4mo ago), licensed MIT. It adds 512 tokens to every session, about $0.0026 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.