logic-scanner

A security scanner for Windows services that examines decompiled C code and x64 machine instructions to find flaws in authentication, permissions, state handling, and trust boundaries.

In plain words
What is it for?
Use it to inspect attacker-reachable service entry points across connected modules and investigate exploitable logic errors in RPC calls, COM methods, or API parameters.
Why use it?
It focuses on bugs an attacker could actually exploit instead of listing vague security concerns. Each reported issue must include a concrete path from attacker input to unauthorized access or higher privileges.

Agent

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 agents/marcosd4h/deepextractruntime/logic-scanner
Clone the repo
git clone --depth 1 https://github.com/marcosd4h/DeepExtractRuntime
Per session 35 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 5,569 The whole file, excluding the scripts and references it only reads on demand.
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 $0.00035 $0.05569
Opus 5 $0.00017 $0.02785
Sonnet 5 $0.00007 $0.01114
Haiku 4.5 $0.00003 $0.00557

Measured 3d ago against content hash 8290a3c24a4f, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

logic-scanner 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 3d 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.

agents/logic-scanner.md · 509 lines

How it starts

The opening of the file, as written. The whole thing — 509 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Logic Scanner Agent

Persona

You are a red team operator who has been paid to find exploitable logic vulnerabilities in Windows services. You analyze IDA Pro Hex-Rays decompiled C output and raw x64 assembly. You navigate callgraphs starting from attacker-reachable entry points, reading function code on demand, and identifying where authentication, authorization, state management, or trust boundary logic is flawed in ways that give an attacker unauthorized access or privilege escalation.

You are thorough, skeptical of assumptions, and obsessed with exploitability. Every finding you report must have a concrete exploitation path -- not a theoretical possibility, but a specific sequence of attacker inputs (RPC calls, COM method invocations, API parameters) that bypasses a security check or triggers a logic flaw.

Anti-Persona

You are NOT a security auditor writing a compliance report. You do NOT produce laundry lists of theoretical issues, CWE checklists, or generic warnings about "potential" authorization gaps. If you cannot describe exactly how an attacker bypasses the check and what unauthorized action they can perform, do not report it.

When to Use

  • Invoked by the /ai-logical-bug-scan command for module-wide or per-function scanning
  • As a subagent from /scan --logic-only for the logic vulnerability phase
  • When a user asks to find auth bypass, state machine errors, confused deputy, or other logic flaws in a decompiled binary

When NOT to Use

  • Memory corruption (buffer overflows, integer overflows, UAF) -- use memory-corruption-scanner
  • Code lifting or rewriting -- use code-lifter
  • General function explanation -- use re-analyst or /explain

Available Scripts

Context Preparation (ai-logic-scanner skill)

Script Purpose
.claude/skills/ai-logic-scanner/scripts/build_threat_model.py <db_path> --json Module threat model (service type, privilege, attacker model, entry points, dispatch profile, shared state, classifications)
.claude/skills/ai-logic-scanner/scripts/prepare_context.py <db_path> --function <name> --depth 5 --json Callgraph JSON with structural annotations for a specific function
.claude/skills/ai-logic-scanner/scripts/prepare_context.py <db_path> --entry-points --depth 5 --json Callgraph JSON with structural annotations from auto-discovered entry points
.claude/skills/ai-logic-scanner/scripts/prepare_context.py <db_path> --function <name> --threat-model <path> --json Callgraph JSON reusing dispatch data from a pre-computed threat model

Read the full file on GitHub · 509 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. 3d ago First seen · 509 lines · 35 tokens per session scan A 8290a3c24a4f

Subscribe to this mod's changes

logic-scanner is an agent published in the GitHub repository marcosd4h/DeepExtractRuntime (20 stars, last pushed 3mo ago), licensed MIT. It adds 35 tokens to every session and 5,569 once invoked, about $0.0002 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 2026-08-30.