# How the PRIMARY Routing System Works in reverse-skill: A 3-Step Guide > Discover how the PRIMARY routing system in reverse-skill works to efficiently match task hints with keywords, loading only essential skills to prevent overload. Learn the 3 steps to optimize your skill paths. - Repository: [ZhaoXu/reverse-skill](https://github.com/zhaoxuya520/reverse-skill) - Tags: deep-dive - Published: 2026-08-07 --- **The PRIMARY routing system in reverse-skill is a fast-track triage mechanism that matches task hints against a curated keyword table to return a single skill path, ensuring only the necessary PRIMARY skill is loaded to prevent "skill-stack overload."** The reverse-skill framework uses a lightweight routing pipeline to determine which skill module should execute for a given reverse engineering or security task. At the heart of this system is the PRIMARY routing mechanism, a deliberate three-step process defined in [`skills/MASTER-ROUTING.md`](https://github.com/zhaoxuya520/reverse-skill/blob/main/skills/MASTER-ROUTING.md) and implemented via `skills/scripts/master-route.ps1`. This architecture guarantees a single-hop, reproducible entry point for every analysis case while maintaining strict isolation between unrelated skill sets. ## The 3-Step PRIMARY Routing Process The PRIMARY routing mechanism operates as a deterministic pipeline that moves from task intake to execution without loading unnecessary modules. ### Step 1: Triage the Task with master-route.ps1 The process begins by identifying the correct skill module for your objective. You can either open [`skills/MASTER-ROUTING.md`](https://github.com/zhaoxuya520/reverse-skill/blob/main/skills/MASTER-ROUTING.md) manually or run the helper script `skills/scripts/master-route.ps1` with a **hint** that describes your target. The PowerShell script parses the hint, matches it against the curated keyword table in [`MASTER-ROUTING.md`](https://github.com/zhaoxuya520/reverse-skill/blob/main/MASTER-ROUTING.md), and outputs a single **PRIMARY** path. This eliminates manual searching through the skill hierarchy. ```powershell # 1️⃣ Quick triage – obtain the PRIMARY skill path .\skills\scripts\master-route.ps1 -Hint "dump Windows NTLM hashes from LSASS" # Expected output (example) # PRIMARY = skills/windows-ad/ # Reason = “Credential Dumping – LSASS” ``` If you run the script without providing a hint, it automatically falls back to the full three-axis matrix in [`skills/routing.md`](https://github.com/zhaoxuya520/reverse-skill/blob/main/skills/routing.md) for manual selection. ### Step 2: Open the PRIMARY Skill and Execute ACTION REQUIRED Once the script returns the path—such as `skills/ida-reverse/`, `skills/windows-ad/`, or the multi-stage `attack-chain/` directory—the framework opens that skill’s [`SKILL.md`](https://github.com/zhaoxuya520/reverse-skill/blob/main/SKILL.md) file. According to the workflow mandates in [`RULES.md`](https://github.com/zhaoxuya520/reverse-skill/blob/main/RULES.md) and [`README.md`](https://github.com/zhaoxuya520/reverse-skill/blob/main/README.md), the system immediately runs the **ACTION REQUIRED** block specified in that document. ```powershell # 2️⃣ Immediately open the skill and run its ACTION block # (the framework does this automatically; in a manual flow you can:) code .\skills\windows-ad\SKILL.md # open the markdown # Follow the ACTION REQUIRED instructions in the file ``` This step ensures that execution begins instantly with the correct contextual instructions, rather than requiring the analyst to navigate multiple documentation layers. ### Step 3: Handle Ambiguity with the Fallback Matrix When the hint does not match any strong keyword in the routing table, [`MASTER-ROUTING.md`](https://github.com/zhaoxuya520/reverse-skill/blob/main/MASTER-ROUTING.md) triggers a deterministic fallback. The system assigns `PRIMARY=R0` and directs you to consult the comprehensive three-axis matrix in [`skills/routing.md`](https://github.com/zhaoxuya520/reverse-skill/blob/main/skills/routing.md) for manual disambiguation. This fallback corresponds to the **"未命中强关键词"** (missed strong keywords) row in [`MASTER-ROUTING.md`](https://github.com/zhaoxuya520/reverse-skill/blob/main/MASTER-ROUTING.md), ensuring that ambiguous tasks still resolve to a valid entry point rather than failing silently. ## Key Files Behind the PRIMARY Routing System The routing logic is distributed across several authoritative source files that define the triage logic, role assignments, and identity models: - **[`skills/MASTER-ROUTING.md`](https://github.com/zhaoxuya520/reverse-skill/blob/main/skills/MASTER-ROUTING.md)** – Contains the primary fast-track table and one-shot script instructions that map keywords to skill paths. - **`skills/scripts/master-route.ps1`** – The PowerShell implementation that executes the triage logic and returns the PRIMARY path string. - **[`skills/routing.md`](https://github.com/zhaoxuya520/reverse-skill/blob/main/skills/routing.md)** – The full three-axis matrix used when hints are ambiguous or when manual override is required. - **[`skills/ops/role-map.md`](https://github.com/zhaoxuya520/reverse-skill/blob/main/skills/ops/role-map.md)** – Defines how the PRIMARY skill is assigned to the *lead* role, who then distributes secondary tasks to other team members. - **[`skills/ops/IDENTITY.md`](https://github.com/zhaoxuya520/reverse-skill/blob/main/skills/ops/IDENTITY.md)** – Documents the three-axis identity model (target type / intent / toolchain) that underpins the routing decisions. - **[`RULES.md`](https://github.com/zhaoxuya520/reverse-skill/blob/main/RULES.md)** and **[`README.md`](https://github.com/zhaoxuya520/reverse-skill/blob/main/README.md)** – Establish the high-level workflow requirements that mandate PRIMARY routing for all engagements. ## Why the PRIMARY Routing System Uses a Lightweight Design The reverse-skill framework deliberately restricts loading behavior to maintain performance and clarity. By design, **only the PRIMARY skill and any explicitly required secondary helpers** are loaded into the execution context. This prevents "skill-stack overload," a condition documented in [`community-security-skills.md`](https://github.com/zhaoxuya520/reverse-skill/blob/main/community-security-skills.md) where extraneous modules clutter the analysis environment and introduce conflicting dependencies. Role assignments from [`skills/ops/role-map.md`](https://github.com/zhaoxuya520/reverse-skill/blob/main/skills/ops/role-map.md) reinforce this isolation. The *lead* role receives the PRIMARY path and maintains responsibility for invoking secondary skills, ensuring that each analysis context remains bounded and reproducible. The three-axis identity model in [`IDENTITY.md`](https://github.com/zhaoxuya520/reverse-skill/blob/main/IDENTITY.md) further refines this by categorizing tasks across target type, intent, and toolchain axes before the PRIMARY fast-path finalizes the selection. ## Summary - The PRIMARY routing system uses `skills/scripts/master-route.ps1` to match task hints against [`skills/MASTER-ROUTING.md`](https://github.com/zhaoxuya520/reverse-skill/blob/main/skills/MASTER-ROUTING.md) and return a single skill path. - Valid PRIMARY paths point to concrete directories like `skills/windows-ad/` or `attack-chain/`, where the [`SKILL.md`](https://github.com/zhaoxuya520/reverse-skill/blob/main/SKILL.md) **ACTION REQUIRED** block executes immediately. - Ambiguous hints trigger a fallback to `PRIMARY=R0` and redirect analysis to the full matrix in [`skills/routing.md`](https://github.com/zhaoxuya520/reverse-skill/blob/main/skills/routing.md). - The architecture prevents skill-stack overload by loading only the PRIMARY skill and necessary helpers, with role distribution managed via [`skills/ops/role-map.md`](https://github.com/zhaoxuya520/reverse-skill/blob/main/skills/ops/role-map.md). ## Frequently Asked Questions ### What happens if the hint doesn't match any keywords in the PRIMARY routing system? The system falls back to a default `PRIMARY=R0` assignment and prompts you to consult the full three-axis matrix in [`skills/routing.md`](https://github.com/zhaoxuya520/reverse-skill/blob/main/skills/routing.md) for manual disambiguation. This corresponds to the **"未命中强关键词"** row in [`skills/MASTER-ROUTING.md`](https://github.com/zhaoxuya520/reverse-skill/blob/main/skills/MASTER-ROUTING.md), ensuring you still reach a valid skill entry point rather than encountering an error. ### How does the PRIMARY routing system prevent skill-stack overload? According to the framework documentation in [`community-security-skills.md`](https://github.com/zhaoxuya520/reverse-skill/blob/main/community-security-skills.md), the system loads **only** the PRIMARY skill and any explicitly required secondary helpers. By avoiding the simultaneous loading of all available skill modules, the framework maintains a clean execution environment free from dependency conflicts or cognitive clutter. ### What is the three-axis identity model in reverse-skill? The three-axis identity model, defined in [`skills/ops/IDENTITY.md`](https://github.com/zhaoxuya520/reverse-skill/blob/main/skills/ops/IDENTITY.md), categorizes every task across three dimensions: **target type**, **intent**, and **toolchain**. This model provides the categorical foundation for the PRIMARY routing table, while the fast-path mechanism in `master-route.ps1` streamlines the selection into a single PRIMARY assignment. ### Can I manually override the PRIMARY routing decision? Yes. You can bypass the automated triage by directly editing [`skills/MASTER-ROUTING.md`](https://github.com/zhaoxuya520/reverse-skill/blob/main/skills/MASTER-ROUTING.md) or by opening [`skills/routing.md`](https://github.com/zhaoxuya520/reverse-skill/blob/main/skills/routing.md) to manually select from the full three-axis matrix. In manual mode, you specify the PRIMARY path yourself before opening the corresponding [`SKILL.md`](https://github.com/zhaoxuya520/reverse-skill/blob/main/SKILL.md) and executing the **ACTION REQUIRED** block.