> For clean Markdown of any page, append .md to the page URL. > For a complete documentation index, see https://docs.method.security/llms.txt. > For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://docs.method.security/_mcp/server. # Continuous External Challenge > Run validated Findings through PoC and exploitation under Rules of Engagement you control. This is the second stage of the Continuous Challenge maturity ladder. [Black Box External Assessment](/guides/best-practices/continuous-challenge/black-box-external-assessment) produces a live inventory and a queue of validated Findings. This stage turns on the autonomous Targeting workflow that takes those Findings through PoC and exploitation, starting with a human in the loop and graduating to Full auto as you build trust. ![Targeting Configuration overview showing five Method Packages (General, Web App, Network, Known Software, Cloud) enabled at Stop after Validation with 100% Trigger coverage and All environments selected.](/_fern-img/d5ccf297a6a3632efdd527ac43270564d0ca423b5950cf65800511aba48afa0d.webp) Turning on Continuous External Challenge will give you: * A daily queue of Compromised or Vulnerable Targets that Method Agents have autonomously produced a working PoC for or compromised, ready for you to review. * A funnel that tracks every validated Finding from Potential Target through Targeted, Validated, Vulnerable, Compromised, and Remediated across every Environment. * All while respecting the Rules of Engagement that you configure and control. Targeting starts conservative, with a human in the loop on every Vulnerable Target. Over weeks, you tune the Rules of Engagement and graduate to Full auto on the surfaces as you build trust in the system. --- ## Setup The Targets app organizes Targeting into Packages. A Package bundles a set of Triggers (the Target Types Method should watch for), an Environment scope, the Rules of Engagement, and the three Agents that run each phase: a Validation Agent, a Pentest Agent, and an Compromise Agent. Method ships with pre-built Packages covering common surfaces: General, Web App, Network, Known Software, Cloud, and Identity. #### Open a Method Package Go to **Targeting > Configuration > Packages**. You will see Method's pre-built Packages listed with their current status and Trigger coverage. Open the **Method General Targeting Package** first. It covers the broadest set of Triggers (API applications, exposed services, SNMP detections, and dozens more) and is the safest place to start. The Package configuration form has four sections: Basics, Triggers, Environments, Rules of engagement, and Agents. ![Top half of the Method General Targeting Package configuration form showing Basics (name and description) and a Triggers list of 48 Target Types including API Application and SNMPv1 Detection.](/_fern-img/de44e4be3f3c949edf127edb656ee10958a083a012ba2bcc949d4e546d762aaa.webp) ![Bottom half of the same Package configuration form, with the All environments toggle on, Rules of engagement set to Stop after Validation, and three Agents assigned: General Validation Agent, General Pentest Agent, and General Compromise Agent.](/_fern-img/7d0ccd180b9484d23c5660b63e0d7c88eb965cbf7d4163951e5a38a34b57d559.webp) #### Set the Rules of engagement to Stop after Pentest In the **Rules of engagement** section, select **Stop after Pentest**. Method will auto-validate matching Findings, auto-run the Pentest Agent to produce a PoC, then wait for human approval before any Compromise Agent runs. This is the recommended starting point the first time you turn on a Package. You get the throughput of autonomous validation and PoC across every matching Finding, with a deliberate gate before exploitation. Leave **All environments** toggled on unless you want to scope this Package to a subset. #### Confirm the assigned Agents Each Package has three Agents assigned by default. The Method General Targeting Package is wired to **General Validation Agent**, **General Pentest Agent**, and **General Compromise Agent**. Open each Agent from the [Agent Fleet](/platform/core-platform/agents) to review its system prompt, MCP Tools, and Skills before you turn the Package on. Swap in a custom Agent here if you have built one for these Triggers. Save the Package. Repeat for any other Method Packages you want to enable on day one. --- ## Refine the guardrails further Packages set the broad strokes, such enabling specific Triggers, which Environments are in scope, and where Method pauses for approval. You also have additional levers to refine where and how Method operates: an exclusion list per Environment, customizations to the validation Agents, and per-Target Type Rules of Engagement on top of the per-Package RoE. Each step below is optional, and they can be completed in any order. #### Define No Strike Lists for each Environment No Strike Lists keep critical or out-of-scope Objects off Method's target list. They apply to every scan, Operation, and Targeting workflow in the Environment. Open the [Administration app](/platform/core-platform/administration), select an Environment, and find the **No Strike List** section on the Overview tab. Add Protection Rules by filter (such as a domain pattern) or by individual Object. The **No Strike List Preview** counter shows how many Objects fall under your rules. Common cases for No Strike entries: * Assets your organization does not own * Customer-facing systems where active probing carries unacceptable risk * Anything explicitly out of scope for an engagement ![Administration app showing the Zamunda Shipping Environment on the Overview tab, with the No Strike List section highlighted in yellow, a filter-based Protection Rule for zamunda-industries.com, and a No Strike List Preview showing 1,436 Objects under protection.](/_fern-img/7f993e50b55c399bf0ee6d8762c55ea3af12488f6b74b71147bac5435adec267.webp) #### Review the Agents in Agent Fleet and add skills Every Target Type has one or more validation Agents that probe matching Findings to confirm exploitability. Method ships with a starter Fleet covering common exposures across cloud, identity, network, and application surfaces. Open the [Agent Fleet](/platform/core-platform/agents) and filter **Agent Type** to **Targeting Validation**. Open any Agent to inspect its system prompt, model, attached MCP Tools, and trigger criteria. Duplicate an Agent to fork it for customization, or click **New Agent** to author one from scratch. Encode your team's tradecraft through the system prompt, the MCP Tools you give the Agent, and SKILL.md files you publish in the **Skills** tab. Govern the Fleet with [Policies](/guides/agents/create-a-policy) to enforce approve, deny, and require-approval rules across Agents, Environments, and MCP Tools. The same review applies to the Pentest and Compromise Agents wired to each Package. ![Agent Fleet on the Agents tab with the Agent Type filter set to Targeting Validation, showing a grid of validation Agent cards covering cloud, identity, network, and application surfaces. Each card lists the Agent's trigger and Tool counts.](/_fern-img/05edced399f6538954bf871cd8607b98769dc822056040035d275b9a6793e098.webp) For background on Agents and Policies, see [AI Agents](/platform/core-platform/agents). For a walkthrough on creating Policies, see [Create a Policy](/guides/agents/create-a-policy). #### Set Rules of Engagement per Target Type Package-level RoE applies uniformly to every Trigger inside a Package. Per-Target Type RoE adds a finer matrix on top: lock a specific Target Type to a more conservative posture (or relax it) regardless of which Package selects it. Open **Targeting**, go to **Configuration**, and select a Target Type from the left rail. On the **Rules of Engagement** tab, set the **Default** RoE. Use **Environment Overrides** to relax automation in lower environments while keeping production conservative. At the top, the **Use as Target** toggle gates whether Findings of this type enter the Targeting funnel at all. ![Targeting Configuration screen for the Cloud Bucket with Open Access Target Type, with the Use as Target toggle enabled, a Default Rule of Engagement of Stop after PoC, and an Environment Overrides dropdown open in yellow showing Full auto, Stop after validation, Stop after PoC, and Manual only.](/_fern-img/51f508eb0a0330615635fa78633af3bb5fab2fbb5b55fd95743957f868c082b0.webp) --- ## What this changes when Method runs The Findings funnel from earlier stages continues to populate. Findings that match an enabled Package's Triggers now enter the Targeting workflow. ### Findings become Targets When a new Finding matches an enabled Package's Triggers, Method opens a Target for it. The Target carries the Finding context, the affected Objects, and the Package that selected it. Targets are visible on the **Target** tab and progress through the funnel: Potential Targets, Targeted, Validated, Vulnerable, Compromised, Remediated. ### The Pentest Agent runs automatically For any Package set to Stop after Pentest, the Pentest Agent runs as soon as the Validation Agent confirms exploitability. It develops a working PoC against the validated exposure and attaches it to the Target. Targets with a completed PoC land in the **Vulnerable** stage of the funnel. ### Vulnerable Targets queue for approval The Compromise Agent does not run until you press **Proceed** on a Target. Each Vulnerable Target arrives in the queue with the Validation Agent's reasoning and the Pentest Agent's PoC attached, so you have what you need to make the call. ### No Strike Lists exclude protected Objects Method skips any Object that matches an Environment's No Strike List. Skipped Objects are recorded with the rule that excluded them, so audits and exception reviews have a trail. ### Validation Agents apply your customizations and Policies Findings route to the validation Agent that matches their Target Type. The Agent runs under its current system prompt, MCP Tools, and Skills, inside the guardrails of the Policies governing the Fleet. Customizations you make in the Fleet take effect on the next matching Finding. --- ## Build trust, then graduate to Full auto We recommend starting the Targeting machine by enabling autonomous Validation and PoC development, followed by manual approval for Compromise Agents to take over. However, over days or weeks of running and reviewing manually, you can tune the RoE per Package and move to Full auto where the Agents have earned trust. **Week one.** * Leave every enabled Package at **Stop after Pentest**. * Check the Target queue daily. * For each Vulnerable Target, read the Validation Agent's reasoning and the Pentest Agent's PoC, then press **Proceed** or **Abort**. * Note which Triggers consistently produce clean validations and PoCs. **Adjust RoE per Package.** * A Package whose Triggers cleanly validate and PoC across the board is a candidate to graduate. * A Package producing noisy or risky PoCs is a candidate to drop back to **Stop after Validation** while you tune the Agents in the [Agent Fleet](/platform/core-platform/agents) or tighten the per-Target Type RoE. **Graduate to Full auto.** * When a Package has earned trust, set its RoE to **Full auto**. * The Compromise Agent runs on every Validated, Vulnerable Target without waiting for approval. * You can still review what was compromised after the fact, but the daily queue shrinks to the Packages you have not yet graduated. **Manual only as a safety valve.** Drop a Package to **Manual only** if you ever want to investigate any Agent behavior or Environment particulars. In Manual mode, every phase, including validation, waits for explicit approval. Use this as a temporary brake while you investigate, not a steady state. --- ## Day-to-day usage The **Targets** app's Priority tab shows Vulnerable and Compromised Targets for your review, and the Target tab shows the entire funnel. ![Target tab showing the funnel (34,135 Potential Targets, 2,269 Targeted, 2,472 Validated, 4 Vulnerable with 1 Blocked, 0 Compromised, 0 Remediated) above a queue of five Targets, each with Proceed and Abort buttons.](/_fern-img/1976b56aad19e09ab478d8d6352649c2a72a46a8faed02463489c0bbdfc24059.webp) * **Work the queue.** Targets in **Vulnerable** that require input list at the top of the tab. Open a Target to see the Validation Agent's findings, the Pentest Agent's PoC, and the affected Objects before pressing **Proceed** or **Abort**. * **Watch the funnel.** The counts at the top show how many Potential Targets are in scope, how many became Targeted, and how many reached each downstream stage. The **Blocked** count surfaces RoE conflicts and No Strike List matches worth reviewing. * **Pivot into [Operations](/platform/core-platform/operations).** When a Vulnerable Target needs a deeper look, jump into Operator. Method seeds the Operation with the Target's context, the Pentest PoC, and the underlying Objects, so you land ready to work. * **Track remediation.** Once the underlying [Finding](/platform/core-platform/findings) is closed, the associated Target moves to **Remediated**. The Target page becomes a record of what was found, compromised, and resolved over time.