Product Appendix 1: Stryct Finder

This Product Appendix supplements the Master Subscription Agreement between Taint Labs Corporation ("Taint Labs", "we") and the Customer named on the Order Form ("Customer", "you"). It applies only to Stryct Finder.

Version
1.0
Published
November 15, 2026
Entity
Taint Labs Corporation

01Scope, structure and order of precedence

1.1 What this Appendix does. This Appendix describes Stryct Finder, the units in which it is sold, and the terms that apply because Stryct Finder works by running adversarial tests against systems Customer controls. It is part of the Agreement. Anything not addressed here is governed by the Master Subscription Agreement ("MSA").

1.2 Order of precedence. If documents in the Agreement conflict, the following order applies (highest first):

  1. Order Form (only for commercial terms it expressly overrides)
  2. Data Processing Addendum (for processing of Personal Data only)
  3. Information Security Addendum
  4. Product Appendix
  5. Master Subscription Agreement
  6. Support and Service Level Policy, Acceptable Use Policy

The Early Access and Pilot Terms ("Early Access Terms") control over all of the above for Early Access Services only. See Section 15 (Early Access status).

1.3 Cross-references. References to sections of the MSA are by section name and number. [●]

02Definitions

Capitalized terms not defined here have the meanings given in the MSA, including "Agreement", "Order Form", "Services", "Stryct Finder", "Customer Agent", "Evaluation" / "Evaluation Run", "Results", "Customer Data", "Attack Catalog", "Third-Party Services", "Users", "Documentation", "Early Access Services", "Personal Data" and "Subprocessor". In this Appendix:

2.1 "Attack Run" means the part of an Evaluation Run in which a test derived from the Attack Catalog is executed against a Customer Agent.

2.2 "Benign Twin" means a run of the same task, in the same setup, with the attack content removed, executed separately alongside its Attack Run so that behavior caused by the attack can be distinguished from behavior the Customer Agent would exhibit anyway.

2.3 "Finding" means an item in the Results reporting that, in an Attack Run, the Customer Agent took an action it was not authorized to take, that the action traced back to untrusted content the Customer Agent read, and that the corresponding Benign Twin did not take the same action, as further described in the Documentation.

2.4 "Integration" means a connection between Stryct Finder and a Customer-designated system, including a source-code repository, a continuous integration ("CI") pipeline, a pull-request check, or the Taint GitHub Action.

2.5 "Runner" means the component that executes Evaluation Runs. A "Hosted Runner" is operated by Taint Labs. A "Customer-Run Runner" is operated by Customer in Customer's own environment under Section 4.

2.6 "Target Environment" means every system, account, dataset, tool, MCP server, API endpoint, mailbox, repository and other resource that a Customer Agent can read from, write to, or act upon during an Evaluation Run.

2.7 "Test Credentials" means any API key, token, OAuth grant, password, certificate or other secret that Customer provides to, or makes reachable by, Stryct Finder or a Customer Agent for the purpose of an Evaluation Run.

2.8 "Stryct Studio" means the Taint Labs web application in which Users manage API keys, review Results and configure the Services. [●]

03Service description

3.1 Overview. Stryct Finder is an evaluation harness that runs attack tests against a Customer Agent before a change is released, and reports what the Customer Agent did. It is delivered as an online subscription service. It performs the functions in Sections 3.2 through 3.6, to the extent those functions are generally available or are enabled for Customer under the Order Form. A function described in the Documentation as "coming soon" or not yet available is not part of the Services until Taint Labs makes it available.

3.2 Discovery. Stryct Finder inspects the Customer Agent, and where an Integration is configured, the associated repository, to identify the models, prompts, tools, MCP servers and authentication scopes that the Customer Agent uses or declares. Discovery from source code finds what the code declares and may run; it does not establish what is actually deployed. [●]

3.3 Attack selection. Based on discovery, Stryct Finder selects tests from the Attack Catalog that are relevant to the tools and data the Customer Agent can access. Stryct Finder may also generate or adapt attack variants for what it discovered, including adaptive variants that change a later payload based on an earlier outcome; adaptive Results are reported separately from fixed ones. [●] Selection is a prioritization, not an exhaustive enumeration (see Section 8.4).

3.4 Attack Runs and Benign Twins. Each Attack Run is paired with a Benign Twin. Both are separate runs with a fresh agent session. Stryct Finder records the inputs the Customer Agent was exposed to, the actions it took, the observable side effects at the tool boundary, and the result of the Benign Twin. Stryct Finder may stop a run when it reaches an execution budget (for example a limit on time, tool calls or tokens). [●]

3.5 Results. Results include, as applicable: Findings; the trace of steps that led to each Finding; the untrusted input the action traced back to; a severity; per-attack and aggregate measurements such as attack success rate ("ASR") and false-interrupt rate ("FIR") with sample sizes and confidence intervals; and a suggested fix. Results are delivered in Stryct Studio and, where an Integration is configured, in Customer's CI run. Metric definitions are those published in the Documentation at the time of the Evaluation Run.

3.6 CI and pull-request integration. Customer may configure Stryct Finder to run from its CI pipeline, using a Stryct Studio API key stored as a repository secret and the Taint GitHub Action, so that Results appear beside the code under review. Runs may be triggered on events Customer configures (for example a merge to the main branch, a release, a pull request, or on demand). Stryct Finder may compare a run with the previous run and re-execute only the tests affected by a change. [●]

3.7 Suggested fixes. A suggested fix is a recommendation. It may be generated with the assistance of automated tools, including language models. Stryct Finder does not apply changes to Customer's code or systems on its own. [●]

04Deployment modes

4.1 Modes. Stryct Finder is offered in two deployment modes: a hosted deployment, using a Hosted Runner under Section 4.2, and an on-premises deployment, using a Customer-Run Runner under Section 4.3. The Order Form states which deployment mode or modes apply.

4.2 Hosted Runner. Where the Order Form specifies a Hosted Runner, Taint Labs operates the Runner. Customer Data is processed in the Taint Labs environment and by Subprocessors as described in the DPA. The Customer Agent and the Target Environment it reaches must be reachable from the Hosted Runner in the manner described in the Documentation.

4.3 Customer-Run Runner. Where the Order Form specifies a Customer-Run Runner:

(a) Taint Labs grants Customer, during the Subscription Term, a limited, non-exclusive, non-transferable, non-sublicensable license to install and run the Runner software, in object-code form, in environments Customer controls, solely to perform Evaluation Runs under this Agreement.

(b) Customer is responsible for the host, network, isolation, patching (where Customer controls the update path), and access controls of the environment in which the Runner operates.

(c) The Runner sends to Taint Labs the data described in the Documentation, which may include Results, traces and operational telemetry. [●]

(d) Components of the Runner or the GitHub Action that are distributed under an open-source license are governed by that license, which controls over this Appendix for those components. [●]

(e) The Information Security Addendum applies to Taint Labs' own systems. It does not apply to the Customer environment in which a Customer-Run Runner operates.

(f) When the Subscription Term for a Customer-Run Runner expires or is terminated, the license in Section 4.3(a) ends, and Customer will stop running the Runner and uninstall it [●]. The end of that license does not affect Customer's ownership of the Customer Data stored in its environment, or its right under MSA Section 8.1 to continue using Results from completed Evaluation Runs it has paid for, and it will not prevent Customer from accessing or exporting that Customer Data or those Results. [●] Taint Labs retains all rights in the Runner, the evaluation harness and the Attack Catalog, and no right to continue using them survives the end of the license. Return and deletion of any Customer Data that Taint Labs itself holds is governed by Section 12.3, separately from this Section 4.3(f).

05Units of measure, entitlements and overage

5.1 Units. The Order Form states the quantity Customer purchases of each applicable unit below. [●]

(a) Customer Agents — each distinct Customer Agent registered for Evaluation. [●]

(b) Evaluation Runs — each Evaluation Run, whether triggered from CI, on demand, or on a schedule. [●]

(c) Repositories — each source-code repository connected through an Integration.

(d) Users — each named User with access to Stryct Studio.

5.2 Usage measurement. Taint Labs measures usage through the Services. Usage data available to Customer in Stryct Studio is the record of usage for billing, absent manifest error. [●]

5.3 Overage. If Customer's usage exceeds the quantities on the Order Form, Taint Labs may (a) invoice the excess monthly in arrears at the overage rate stated on the Order Form, or if none is stated, at the per-unit rate on the Order Form pro-rated for the remainder of the Subscription Term; or (b) where the Order Form says so, stop starting new Evaluation Runs until the next period or until Customer purchases additional quantity. Taint Labs will notify the Customer billing contact before invoicing any overage for the first time, and will not stop an Evaluation Run already in progress solely because of overage. [●]

5.4 No carry-over. Unused quantities do not carry over to a later period or Subscription Term unless the Order Form states otherwise.

06Customer responsibilities for adversarial testing

Stryct Finder deliberately tries to make a Customer Agent do things it should not do. The obligations in this Section are a condition of using it.

6.1 Authorization to test. Customer represents and warrants that, for each Evaluation Run, Customer owns, or has current written authorization from the owner to test, the Customer Agent and every part of the Target Environment, within the scope of that authorization. That includes any Third-Party Service the Customer Agent calls, to the extent the test sends attack content to it or causes the Customer Agent to act on it. Taint Labs relies on this representation and does not verify it. Taint Labs cannot authorize Customer to test any system Taint Labs does not control.

6.2 Non-production by default. Customer will run Evaluations only against non-production environments, or against environments whose owners and affected users have consented to adversarial testing, unless a Testing Authorization and Rules of Engagement signed by both parties names the production systems in scope and any additional safeguards (MSA Section 5.4). Customer will populate test environments with synthetic or test data wherever practicable, and will not include real Personal Data, customer records or live secrets in a Target Environment unless that is necessary for the test and permitted by the DPA.

6.3 Test Credentials and least privilege. Customer will:

(a) issue Test Credentials that are dedicated to testing, scoped to the minimum permissions the Evaluation requires, limited to the Target Environment, and revocable;

(b) not provide credentials for production accounts, personal accounts, administrative or organization-wide scopes, or payment instruments, except as permitted under Section 6.2;

(c) store Test Credentials used by an Integration only in the secret store of its CI system or as the Documentation directs, and never in source code or in prompts; and

(d) rotate or revoke Test Credentials promptly after an Evaluation, and at once if Customer suspects exposure.

Some attacks in the Attack Catalog are designed to make an agent disclose secrets it can reach. Customer should therefore treat any secret reachable by a Customer Agent during an Evaluation as potentially exposed in the Results and traces, and choose Test Credentials accordingly. [●]

6.4 Side effects of Customer Agent actions. Customer Agents act through their own tools and credentials. If a Customer Agent sends an email, posts a message, modifies a file, calls a payment or refund API, changes a record, or takes any other action during an Evaluation Run, that action is taken by Customer's system in Customer's environment. As between the parties, Customer is responsible for those actions and their consequences, including actions induced by attack content delivered by the Services, except to the extent the action resulted from Taint Labs conducting the Evaluation outside the Evaluation Scope or otherwise in breach of the Agreement (the same standard as MSA Section 6.2).

6.5 Sandboxing and containment. The Services execute attacks inside a contained harness, but a Customer Agent's own tools reach whatever Customer has connected them to. Customer is responsible for isolating the Target Environment so that actions taken during an Evaluation cannot reach real recipients, third parties, production data, or systems outside the scope of Section 6.1. This includes, where relevant, using test mailboxes and sinkholed outbound destinations, disabling or stubbing irreversible actions (payments, deletions, external messages), and setting spend limits.

6.6 Spend and rate limits. Some tests (for example resource-exhaustion or "denial of wallet" tests) are designed to make an agent consume resources. Customer is responsible for setting spend, rate and quota limits on its own accounts with Third-Party Services.

6.7 Third-party terms. Customer is responsible for complying with the terms of each Third-Party Service the Customer Agent uses during an Evaluation, including any provision of those terms that restricts security testing, automated traffic, or the submission of adversarial content.

6.8 Attack content. Attack content delivered by the Services is intentionally malicious in form (for example injected instructions, poisoned tool descriptions or crafted documents). Customer will not move attack content from a Target Environment into a production system, publish it where a real person's agent could encounter it, or leave it in a shared data store after the Evaluation.

6.9 Personnel. Customer will make sure Users who configure Evaluations understand this Section 6.

07Third-Party Services and consumption costs

7.1 Customer Agent consumption. The Customer Agent's calls to models, APIs, MCP servers and other Third-Party Services during an Evaluation Run (including the Benign Twin) are made with Customer's accounts and credentials and are paid for by Customer, unless the Order Form states otherwise. Customer acknowledges that each Evaluation Run executes many Attack Runs and Benign Twins and that its Third-Party Service charges scale with that number.

7.2 Taint Labs consumption. Any model or service that Taint Labs itself uses to operate Stryct Finder (for example to select or generate attacks, classify outcomes, or draft suggested fixes) is paid for by Taint Labs and included in the Fees, unless the Order Form states otherwise (MSA Sections 6.1 and 6.4).

7.3 No responsibility for Third-Party Services. Third-Party Services are not part of the Services. Taint Labs is not responsible for their availability, behavior, pricing, or changes to them, including a change that alters a Customer Agent's Results between runs.

08Results: nature and limitations

8.1 Results are measurements. Results are point-in-time measurements of a specific configuration of a Customer Agent — its model version, prompts, tools, permissions, policy and runtime as observed in the Evaluation Run -- against a specific set of tests, with a stated number of trials. They say what happened when that configuration met those tests. They do not describe behavior in any other configuration or environment.

8.2 Probabilistic. The systems Stryct Finder tests are non-deterministic. The same test against the same configuration can produce different outcomes on different runs. Results are reported with sample sizes and, where applicable, confidence intervals, and are subject to sampling error. Results include runs that could not be decided (for example because a run errored, exhausted its budget, or the attack never reached the model); those runs are reported and excluded from rates, and an undecided run is not a pass.

8.3 How outcomes are determined. Stryct Finder determines whether an attack succeeded primarily from observed effects — what the tools, fixtures and endpoints in the harness actually recorded — checked against fixed authorization rules and the provenance of the content the Customer Agent read, and compared with the Benign Twin. Some measurements, severity assignments, explanations and suggested fixes may rely on automated classifiers, including language-model-based judges, which can be wrong. [●]

8.4 No guarantee of coverage. The Services test for the attack techniques in the Attack Catalog that Stryct Finder selects for a Customer Agent. They do not test for every possible attack, every variant of an attack, or attacks not yet known. A Customer Agent with no Findings may still be vulnerable. A Finding identifies one path; it does not imply that it is the only one.

8.5 Not a certification. Results, and any score or grade in them, are not a security certification, accreditation, audit opinion, attestation, seal, or evidence of compliance with any law or standard, and are not an endorsement. A grade is the output of the rubric published in the Documentation applied to the run's data; it is Taint Labs' opinion on that data. Customer will not represent Results to any third party as any of those things. Customer may share Results with its auditors, regulators, insurers, customers and prospective customers under its own name, provided Customer presents them as measurements of a named configuration with their date, sample size and caveats, and does not state or imply that Taint Labs certified or endorsed the Customer Agent.

8.6 Caveats travel with the Results. Results may carry caveats (for example that a metric is a proxy rather than ground truth, or that a check was a consistency check rather than a replay). Those caveats are part of the Result. Customer will not remove them when sharing Results outside its organization.

8.7 False positives and false negatives. Results may contain false positives and false negatives. Customer is responsible for reviewing Findings, validating them in its own environment, and deciding whether and how to remediate.

8.8 Suggested fixes. Customer is responsible for reviewing and testing any suggested fix before applying it. A suggested fix may not resolve the Finding, may resolve it only for the tested attack, and may affect the Customer Agent's functionality.

8.9 Methodology and catalog changes. The Attack Catalog, the rubric, severity weights and metric definitions change over time. Results from different catalog, rubric or harness versions may not be comparable, and a re-run of an unchanged Customer Agent may produce different Results because the tests changed. Each Result records, where available, the versions used to produce it. [●]

8.10 Customer remains responsible for its security. The Services are one input to Customer's security program. They are not a substitute for Customer's own testing, threat modeling, code review, access control and runtime controls.

09Attack Catalog license and restrictions

9.1 License. During the Subscription Term, Taint Labs grants Customer a limited, non-exclusive, non-transferable, non-sublicensable license to use the Attack Catalog, and the attack content, payloads and test definitions delivered through the Services, solely to evaluate Customer Agents that Customer is authorized to test under Section 6.1, for Customer's internal business purposes. This license includes the right to reproduce a specific Finding inside Customer's own environment to confirm it and to verify a fix, including by keeping a regression test for that Finding in Customer's own code.

9.2 Restrictions. Customer will not, and will not permit anyone to:

(a) use the Attack Catalog or any attack content against any system Customer does not own or is not authorized in writing to test;

(b) host, plant or publish attack content anywhere a real person's agent could encounter it;

(c) extract, export, bulk-download or systematically copy the Attack Catalog or attack content delivered through the Services, except the content of Customer's own Results;

(d) use the Attack Catalog or attack content to build, train, fine-tune, distill or evaluate a product or service that competes with Stryct Finder;

(e) resell, sublicense, or provide the Attack Catalog or the Services to a third party, or use them to provide testing services to a third party, unless the Order Form expressly permits it. [●]

9.3 Public catalog. The portions of the Attack Catalog published on docs.stryct.ai remain available to Customer under the public Terms of Service at stryct.ai, independent of this Appendix. Nothing in this Appendix narrows what those terms already allow.

9.4 Ownership. The Attack Catalog, the harness, the rubric, and the Services are Taint Labs' property as set out in the "Intellectual Property; Feedback" section of the MSA (Section 13). Customer's Results are Customer Data. Taint Labs retains its rights in the attack content, templates and generic remediation guidance that appear within Results.

10Publication and the public leaderboard

10.1 No publication without written opt-in. Taint Labs will not publish, display on its public leaderboard, cite in marketing, or otherwise disclose to the public any Results, any Customer Agent's name or configuration, or the fact that Customer uses Stryct Finder, unless Customer has given express, written opt-in for that specific use. Opt-in must identify the Results or information concerned and the intended use, and may be withdrawn for future use by written notice, after which Taint Labs will remove the material from its own properties within a reasonable time, not more than the number of days stated in MSA Section 8.4(c). For the public leaderboard, MSA Section 8.4 governs: Customer receives a preview of the row and correction of factual errors, but no right to approve its content.

10.2 Public evaluations of public models. Taint Labs independently evaluates publicly available models, agents and SDKs and publishes those results. Customer's subscription does not restrict that work, provided Taint Labs does not use Customer Data or Customer's Confidential Information in it. If Customer's Customer Agent is built on a public model that also appears on the leaderboard, the leaderboard row is Taint Labs' own evaluation, not Customer's Results.

10.3 Aggregated data. Any use by Taint Labs of aggregated or de-identified data derived from the Services, including cross-customer attack success statistics, is governed by MSA Section 7.5 and DPA Section 3.4, including their publication limits and Customer's publication opt-out. Any such use will not identify Customer or a Customer Agent.

11Findings in third-party products: coordinated disclosure

11.1 Scope. An Evaluation may reveal that a weakness lies not in Customer's own code or configuration but in a Third-Party Service — for example a model, agent SDK, MCP server or tool supplied by another vendor. This Section governs such a "Third-Party Vulnerability" (as MSA Section 9.1 defines that term).

11.2 Customer's own weaknesses stay confidential. Findings about Customer's own Customer Agent, code and configuration are Customer's Confidential Information and Customer Data. Taint Labs will not disclose them to any third party, including the vendor of a component Customer uses, except as the "Confidentiality" section of the MSA (Section 12) permits.

11.3 Coordinated disclosure. For a Third-Party Vulnerability:

(a) Customer may report it to the affected vendor itself, under that vendor's own disclosure policy.

(b) Taint Labs may, with Customer's written consent, notify the affected vendor on Customer's behalf or jointly. Taint Labs will not identify Customer to the vendor without Customer's consent.

(c) Separately, Taint Labs may report the underlying attack technique to the affected vendor, and may reproduce it in Taint Labs' own environment, in each case only in a form that contains no Customer Data or Customer Confidential Information and does not identify Customer.

11.4 Embargo. Neither party will publish a Third-Party Vulnerability of critical or high severity before the affected vendor has been notified or has expressly waived notification, and each party will give the vendor a reasonable window to remediate before publication. A request for an embargo will not be used to require silence indefinitely. If either party knows that a Third-Party Vulnerability is already under another embargo, it will tell the other and will not publish into it. This approach is consistent with Taint Labs' published vulnerability disclosure practice (stryct.ai/report and the Documentation). [●]

11.5 Attack Catalog additions. If a Third-Party Vulnerability reflects an attack technique not in the Attack Catalog, Taint Labs may add the technique to the Attack Catalog after the embargo in Section 11.4 ends, in a form that contains no Customer Data, does not identify Customer, and credits Customer only if Customer asks for credit.

11.6 Vulnerabilities in Stryct Finder. Customer should report suspected vulnerabilities in Stryct Finder or other Taint Labs systems to security@taint.ai. Testing the Services' own security is governed by the Acceptable Use Policy and the Information Security Addendum, not by Section 6. [●]

12Customer Data in Stryct Finder

12.1 What Stryct Finder stores. Customer Data processed by Stryct Finder may include Customer Agent configurations, prompts, tool and MCP server definitions, repository contents read during discovery, traces of Attack Runs and Benign Twins, tool inputs and outputs, and Results. Traces may include whatever data the Customer Agent read or emitted during a run, including any Test Credentials or Personal Data reachable in the Target Environment (see Sections 6.2 and 6.3).

12.2 Repository access. Where an Integration connects a repository, Stryct Finder reads the repository to the extent needed for discovery and change tracking. [●]

12.3 Retention. Taint Labs retains traces for [●] and Results for [●] after the Evaluation Run, or for the Subscription Term if longer, unless Customer deletes them earlier through Stryct Studio or by written request. On expiry or termination of the Agreement, Customer Data is returned and deleted as set out in the "Data export and deletion" section of the MSA (Section 11.5) and the DPA (Section 13): for a hosted deployment, 30 days of read-only export access, then deletion from production systems within the following 30 days; for an on-premises deployment, those obligations apply to the Customer Data Taint Labs itself holds, and Section 4.3(f) governs the end of the Runner license. [●]

12.4 Model training. Taint Labs will not use Customer Data, and will not permit any Subprocessor or other third party to use Customer Data, to train, fine-tune or otherwise improve any machine-learning model, whether the model is operated by Taint Labs or a third party and whether or not it is made available to anyone other than Customer, as MSA Section 7.4 provides.

12.5 Third-Party Services Customer chooses. Data that a Customer Agent sends to a Third-Party Service during an Evaluation is sent by Customer's system to Customer's provider. That provider is not a Subprocessor of Taint Labs.

13Run safety and suspension

13.1 Stopping a run. Customer may stop an Evaluation Run at any time. [●]

13.2 Suspension for unsafe runs. In addition to any suspension right in the MSA (Section 18), Taint Labs may immediately stop an Evaluation Run, and suspend Customer's ability to start new ones, if Taint Labs reasonably believes that the run:

(a) is reaching, or is likely to reach, a system, person or account outside the scope authorized under Section 6.1;

(b) is targeting a system Customer is not authorized to test;

(c) is causing, or is likely to cause, harm to a third party, to Taint Labs' systems, or to other customers;

(d) is consuming resources in a way that threatens the stability of the Services; or

(e) would violate applicable law or the Acceptable Use Policy.

13.3 Notice and restoration. Taint Labs will notify Customer promptly, and where practicable before acting, will limit the suspension to the runs or Customer Agents concerned, and will restore access once the cause is resolved. A suspension under Section 13.2 does not relieve Customer of Fees, unless the suspension resulted from Taint Labs' error.

13.4 No duty to monitor. Taint Labs' right to stop or suspend a run does not create a duty to monitor Customer's Evaluations or to prevent the side effects described in Section 6.4.

14Product-specific disclaimers

14.1 In addition to the disclaimers in the "Warranties and Disclaimers" section of the MSA (Section 14), and to the maximum extent permitted by law:

(a) TAINT LABS DOES NOT WARRANT THAT TAINT FINDER WILL DETECT ALL OR ANY PARTICULAR VULNERABILITY, THAT A CUSTOMER AGENT WITHOUT FINDINGS IS SECURE, THAT RESULTS ARE FREE OF FALSE POSITIVES OR FALSE NEGATIVES, OR THAT RESULTS ARE REPRODUCIBLE OR PREDICTIVE OF THE BEHAVIOR OF ANY SYSTEM IN ANY OTHER CONFIGURATION, ENVIRONMENT OR TIME;

(b) TAINT LABS DOES NOT WARRANT THAT ANY SUGGESTED FIX WILL REMEDIATE A FINDING OR WILL NOT AFFECT THE FUNCTIONALITY OF A CUSTOMER AGENT; AND

(c) TAINT LABS IS NOT RESPONSIBLE FOR ACTIONS TAKEN BY A CUSTOMER AGENT DURING AN EVALUATION RUN, OR FOR THEIR CONSEQUENCES, EXCEPT AS SECTION 6.4 PROVIDES.

14.2 Customer indemnity for unauthorized testing. Claims arising from Customer's breach of Section 6.1, 6.2, 6.8 or 9.2(a) or (b) are claims for which Customer indemnifies Taint Labs under the "Indemnification" section of the MSA (Section 15.4).

15Early Access status

15.1 Current status. As of the date of this Appendix, Stryct Finder is an Early Access Service. Until Taint Labs designates Stryct Finder generally available in writing (including by notice to Customer or by removing the early-access label from the Documentation and product pages), the Early Access Terms apply to Stryct Finder and control over this Appendix where they conflict. Whether a paid Stryct Finder subscription sold before general availability carries the MSA Section 14.2 warranty, the Support and Service Level Policy or the MSA Section 15.1 indemnity is stated in the Order Form (MSA Section 4.6); absent that statement, none of them applies.

15.2 Effect. While Stryct Finder is an Early Access Service, the Support and Service Level Policy applies only to the extent the Early Access Terms or the Order Form say it does, and features described in this Appendix may be unavailable, limited or changed without the notice required for generally available Services.

15.3 On general availability. When Stryct Finder becomes generally available, this Appendix applies without the Early Access Terms for the remainder of the Subscription Term, except that any feature still labeled early access, beta, preview or pilot remains an Early Access Service. [●]

16Changes to Stryct Finder

16.1 Taint Labs may update the Attack Catalog, the harness, the rubric, and the metrics as part of operating Stryct Finder, and Customer acknowledges under Section 8.9 that this can change Results. Taint Labs will not make a change that materially reduces the overall functionality of Stryct Finder during a paid Subscription Term for a generally available Service, except as the MSA permits. Changes to the Documentation's metric definitions or rubric will be reflected in the Documentation. [●]

17Survival

Sections 4.3(f), 6.4, 8, 9.2, 9.4, 10, 11, 12.3, 14 and this Section 17 survive expiry or termination of the Agreement, together with any other provision that by its nature should survive.