# Praxenor | scoped evaluation planning worksheet Revision 1.0 ยท October 4, 2026 This is a blank planning template, not a product test result, certification, integration specification, commercial offer or authorization to access any system. ## 1. Agree the workflow Product and exact build: ____________________ Responsible reviewer: ____________________ Question to answer: ____________________ Permitted environment and sample source: ____________________ Systems and actions outside the agreed scope: ____________________ Keep credentials, customer data, personal records and suspicious files out of the public website form. Agree on a suitable exchange method separately. ## 2. Record the setup before testing | Requirement | Agreed setup or unresolved question | |---|---| | Covered actions or supported record format | | | Who grants access and approves actions | | | Input origin, timestamps, gaps and authenticity limits | | | Storage location, retention, deletion and backup | | | External requests, exports and permitted recipients | | | Stop conditions and rollback/recovery owner | | | Support arrangements and any commercial terms | | ## 3. Predefine the tests ### Agent Defender: review of permissions for connected actions Use a harmless, explicitly connected workflow. The published brief describes a disposable local Git project, not general email or computer coverage. | Case to agree and test | Expected decision/result | Observed result and receipt | |---|---|---| | Normal action within granted scope | | | | Action without required approval | | | | Request changed after approval | | | | Repeated use or expired authority | | | | Checkpoint unavailable or path bypassed | | | | Save/reopen record and recover after a refusal | | | ### Swarm Hunter: investigation from permitted records Use supported records whose source and limits are documented. Findings are advisory; do not equate a pattern with intent or AI involvement. | Case to agree and test | Expected interpretation | Observed result and record links | |---|---|---| | Ordinary authorized activity | | | | An agreed issue represented in the input | | | | Related activity with a plausible benign explanation | | | | Missing timing, permission or provenance | | | | Save/reopen/export with source labels and assessment | | | ## 4. Keep evaluation separate from development Acceptance thresholds agreed before results: ____________________ Fresh cases kept separate from examples used for development: ____________________ Record sample size, useful findings, false warnings, missed issues, unresolved cases, reviewer disagreement, task completion and reviewer effort. Do not silently discard incomplete cases. Repeated runs of one synthetic case do not create independent evidence. ## 5. Decide and document Result: proceed / revise / stop / unresolved Evidence supporting the decision: ____________________ Known coverage and sample limitations: ____________________ Next authorized step and owner: ____________________ ## Source scope * Current company reported product status: https://praxenor.io/evidence/#current-status * Agent Defender technical brief, revision 1.0, October 3, 2026: https://praxenor.io/veristade/agent-defender/technical/ * Swarm Hunter technical brief, revision 1.0, October 3, 2026, describing build 0.6.0: https://praxenor.io/veristade/swarm-hunter/technical/ * Proposed investigation evaluation: https://praxenor.io/evaluation/ The published status distinguishes local Swarm Hunter 0.6.1 from staged 0.6.2. Confirm the specific build before relying on any described capability. A completed website worksheet does not activate software or authorize an evaluation.