# Paladin in-game test checklist Prepared **7 November 2026** for World of Warcraft: Forever. **This is a test plan, not a completed test or evidence of successful gameplay.** All 36 current Paladin designs remain `researched` and untested in game; the three historical records remain `deprecated`. All 39 canonical `tested_builds` arrays are empty. Use this plan with the [39-record case list](PALADIN_MACRO_TEST_CASES.md), the [blank test record](templates/PALADIN_TEST_RECORD.md), and the existing [validation matrix](VALIDATION_MATRIX.md). The case list identifies each exact macro ID/version and its special checks. Do not substitute a similar macro. ## 1. Choose and freeze a small test batch - [ ] Select current designs from the [canonical pack](../macros/paladin/pack.json). Keep the three deprecated records out of the routine acceptance batch. - [ ] Back up the existing macros and bindings before replacing anything. Use `/macro` for manual creation, or the optional importer with its own documented setup; record which method was used. - [ ] Choose only the alternatives needed for the session. There are 36 current designs; the project's configured character-slot warning begins above 30. Record the actual client's capacity rather than assuming that configuration proves it. Test in batches without overwriting unrelated macros. - [ ] Record each ID, version, in-game name, exact body, UTF-8 byte count and SHA-256 of the UTF-8 body. Keep the source revision and raw pack hash too. Document any newline normalization used in comparisons. - [ ] Compare the body read back from the native macro editor with the selected canonical text. Verify no truncation, parser error or accidental extra lines. Do not append debug Lua or change targets and then attribute results to the original body. A personal edit is a different test subject. This plan's source snapshot is application commit `f5d4822758cc022f22279eb9b04e11e5b08c97ed`, raw Paladin pack SHA-256 `70bb134481befab9747905ea351e651814a850d5aca7c5efbd79713c8090fec7`. If the pack changes, review the affected IDs, versions, bodies and cases before reusing this checklist. This hash identifies content, not a game client build. ## 2. Capture the actual client and character - [ ] Start a real session record with actual date/time, timezone, tester reference, client version, build number, interface number, locale, installation or ruleset, and release/beta/release-candidate channel evidence. - [ ] Record class, level, race/faction, actual talents and Legacy allocation, learned spell names/ranks, weapon/shield, relevant gear and reagents. Capture the spellbook and legal talent allocation. A spec label alone is insufficient. - [ ] Record addon names/versions, action-bar owner, binding and modifier conflicts, self-cast/focus-cast settings, and button-down versus button-up behavior. Test the real intended binding as well as checking the editor. - [ ] If using [Macro Lab](../addon/AllegedlySkilledMacroLab/README.md), record its version, run `/aslab scan`, then `/reload` or log out/exit normally to flush the account-level `SavedVariables/AllegedlySkilledMacroLab.lua`. Preserve the scan ID and original file. Verify the active Forever installation through the launcher; do not assume the historical `_classic_beta_` example is current. - [ ] If not using the collector, preserve equivalent build, spellbook and talent evidence from the actual client. Missing API values remain unknown. The current research export's base **1.60.1.70755** is not a tested build. Never copy it into the session unless that is independently the observed client. The collector describes environment and visible availability; it does not record macro execution. Optional `/asm status` diagnostics and an import receipt also do not prove a successful cast or recipient selection. ## 3. Prepare controlled recipients and ordinary spell controls - [ ] Arrange a living friendly mouseover **A** and a different friendly current target **B**, plus controlled hostile recipients when required. Use actual eligible dead players, party/raid members or creature types for those cases. - [ ] Start with known health/mana, cooldowns, GCD, auras, seal, attack state, range, line of sight, facing and combat state. Reset these between repeats. - [ ] Put the ordinary learned ability from the spellbook on a separate button. Under the same conditions, use it as a control for spell eligibility, rank, targeting and native attack/cast behavior. A failed control identifies a setup/ability question; it does not validate the macro's positive branch. - [ ] Use deliberate single presses. Keep targets, cursor/frame and modifiers visible in the evidence. Do not use combat automation or execute stored Lua. Use a controlled location and cooperating players for crowd control, immunity, damage transfer and threat observations. Do not infer a taunt, immunity or cooldown relationship merely from an icon or successful button activation. ## 4. Run B: baseline for every current macro - [ ] **B1 — identity:** complete the exact-body/native-editor readback above. - [ ] **B2 — positive action:** with the required spell/talent/equipment learned and ready, observe one successful intended action, correct recipient, actual rank where observable, and effect. Compare the ordinary spell control. - [ ] **B3 — presentation:** check icon and tooltip for each relevant modifier branch. A tooltip is presentation evidence, not execution evidence. - [ ] **B4 — selection:** test applicable living friendly, living hostile, dead, absent and conflicting mouseover/target states using the groups below. Record whether the current target changes. Focus/pet branches absent from a body are N/A with a reason; do not invent them. - [ ] **B5 — unusable first match:** after a unit satisfies the authored condition, test relevant range, line-of-sight, recipient eligibility, cooldown/GCD, resource, equipment and unavailable-spell cases. Record the error and recipient; a failed first cast must not be counted as a successful fallback. - [ ] **B6 — side effects:** test applicable stopcasting, stopattack, aura, threat, reagent, channel and combat restrictions in the per-record cases. For melee actions, compare native and macro behavior from attacks stopped, after target changes, on successful casts and on relevant failures. Absence of `/startattack` does not establish native attack behavior. - [ ] **B7 — persistence:** after `/reload`, read back the exact macro and repeat meaningful positive and negative cases. Record normal exit/restart readback if performed. Persistence or a successful import alone is not a gameplay pass. For every required case, record **two consistent observations after resetting the conditions**. This is the practical evidence gate adopted by this checklist, not a statistical guarantee or an existing validator requirement. Long cooldowns can be checked across sessions with separately recorded builds/setups. If a repeat cannot be completed, explain it and retain incomplete coverage for review. ## 5. Run the shared branch groups assigned in the case list These are authored branch expectations to verify, not claims of observed client success. Each group includes B and its per-record special checks. | Group | Cases to check, with distinct recipients A and B | | --- | --- | | **S — Shift support** | Shift selects self even with friendly A/B present. Without Shift, living helpful mouseover A precedes friendly target B. Hostile/dead/missing mouseover permits friendly target B; no eligible friendly recipient reaches the explicit self fallback. Repeat with a matching but unusable A and verify no cast on B/self. | | **A — Alt support** | The same cases using Alt for self. Check Shift alone against the ordinary priority and the real binding settings; do not assume the older Shift convention applies. Verify matched-but-unusable A does not redirect. | | **H — hostile mouseover** | Living hostile mouseover A precedes hostile target B; hostile A works without a target. Friendly/dead/missing mouseover permits an eligible hostile current target. No eligible hostile unit produces no authored cast. Unusable but matching hostile A must not redirect to B. Swift Judgement requires Shift throughout H; its unmodified press has no cast. | | **P — healing-first hybrid** | Alt self, then living friendly mouseover, friendly current target, hostile current target, otherwise self healing. Enemy mouseover alone is ignored and reaches self healing; enemy mouseover A with target B uses the applicable B branch. Failed healing must not become an attack. Verify the priority with simultaneous friendly and hostile units. | | **C — cast/cancel** | Unmodified press attempts the self aura. Shift removes only that existing own aura; repeat Shift and Shift with no aura must not reapply it. Observe another unit's aura remains untouched. Test unavailable/cooldown/restriction states and ensure cancelling an immunity does not imply clearing Forbearance. | Holy Shock, Light's Vigil, Redemption, Sacrifice, seal selection, the dedicated BoP cancel key and the selected-target Holy Strike have distinct bodies. Follow their explicit case-list rows instead of forcing them into a shared group. ## 6. Complete the Paladin-specific and cross-record checks - [ ] **Emergency interruption:** test Hammer of Justice's conditional stopcast and Protection/Lay on Hands/Bulwark's unconditional or effective-every-press cancellation, including failed attempts. Record cancellation before the result. - [ ] **Support effects:** verify actual dispellable effects; blessing replacement; resurrection and combat exclusion; party/class/reagent effects of Greater Blessings; Sacrifice's eligibility and actual transfer. Missing helpers, reagents or effects leave required cases blocked, not passed. - [ ] **Defensives and control:** observe cancel versus cast behavior, Forbearance, immunity restrictions, stopattack on failed control attempts, recipient types and existing periodic damage. Threat claims require actual threat/recipient observations; no coefficient is established merely by a visible aura. - [ ] **Talents, seals and cooldowns:** verify actual acquisition and legal allocation, Fury versus other/no seal, Swift versus ordinary Judgement in both cooldown directions after the actual GCD, and Holy Shield's equipment and real-channel case. Do not invent a channel if none is available. - [ ] **Thresholds and ranks:** verify the actual learned rank and Hammer of Wrath's eligibility. Rounded health percentages do not prove an exact threshold; a heal/damage amount alone does not establish which rank was cast. - [ ] **Vigil recipient alignment:** run the detailed cross-record experiment in the case list. With enemy mouseover A and friendly target B, Vigil's authored branch selects B; standalone Holy Shock selects A, while the healing-first hybrid selects B. These are separate presses. Record actual setup/follow-up effects and changed/lost recipients; neither key remembers the Vigil recipient. - [ ] **Unknown self eligibility:** capture deliberately selected self for Vigil and Sacrifice as diagnostic questions. A helpful condition can attempt self; neither absent fallback nor source data proves eligibility. Keep unknown self-support outside any supported claim until evidenced and reviewed. The case list contains all special requirements for the **36 current IDs**. Keep the three historical records in a separate acquisition investigation. Direct Divine Favor success does not validate the deprecated Voice of Truth selector. A surviving old spell does not automatically reinstate its old macro. ## 7. Save an evidence bundle and resolve incomplete cases - [ ] Complete the [blank form](templates/PALADIN_TEST_RECORD.md) for each exact macro, or share a session environment record while linking each macro's body and case observations unambiguously. Write expectations before execution. - [ ] Map each result to original screenshots, video/sequence or combat-log excerpts with IDs, timestamps/timecodes and file hashes. Preserve state before the press, the single input, actual recipient/action/rank, errors and side effects. A screenshot of the editor proves text, not correct execution. - [ ] Retain originals privately, for example in the existing ignored `research/captures/private/`. SavedVariables can contain names, realms and other identifiers; treat Lua as data and never execute it. Keep private captures out of `public/`, canonical notes and source control. - [ ] Prepare a reviewed, redacted report, for example `research/client-tests/paladin/.md`, with non-identifying evidence references/hashes. Preserve originals so a maintainer can inspect the actual evidence. Do not publish unrelated chat or account dumps. - [ ] Record the actual outcome for every required case using this vocabulary: | Result | Meaning | | --- | --- | | `PASS` | Observed positive behavior matches the stated expectation. | | `EXPECTED_FAILURE` | A predeclared negative case fails in the expected way; this does not prove the positive branch works. | | `FAIL` | Wrong recipient/action, unexpected side effect or another mismatch. | | `BLOCKED` | A required talent, rank, recipient, reagent, channel or other setup is unavailable. | | `NOT_RUN` | No observation was made. | | `NA` | Truly irrelevant to this exact body, with a written reason. Missing prerequisites are not N/A. | Unexplained differences, blocked advertised branches and unfinished cases keep the macro untested for that build. A quick smoke test or one successful branch is partial evidence. Do not put a partially tested new build into the whole macro's `tested_builds` array. Qualify observations by the actual locale, level, known ranks, talents, gear and settings; another setup is not implicitly tested. ## 8. Maintainer evidence review before a status change The report must identify the exact body/version and actual build/channel; contain the applicable baseline, branch, negative and special-case evidence and reset repeats; resolve failures; justify N/A rows; retain inspectable private originals and a safe report; and state the supported scope and remaining limitations. A maintainer reviews this bundle before any separate canonical edit. | Proposed state | Evidence needed | | --- | --- | | Keep `researched` / visibly untested | Source review only, partial gameplay observations, unknown build/channel, or any unresolved required case. Record partial work in the report. | | `beta-tested` | Complete, successful applicable coverage on an evidenced **actual beta build**, with body/version, report and maintainer review. This label cannot stand in for testing on release. | | `release-ready` | Complete applicable retest on the **released client or an explicitly accepted release candidate**, with actual build, sources, evidence and maintainer/release gates satisfied. A release-only session need not invent a beta history; retain researched status until the release retest/review is complete. | | Keep `deprecated` | Historical acquisition or isolated surviving behavior is research evidence. Restoration requires a separate content/design review, not an automatic tested badge. | There is no generic `tested` status. The existing validators chiefly check a nonempty `tested_builds` list for tested statuses (and sources for release-ready); they do **not** verify report authenticity, exact-body coverage, beta/release channel, repeat observations or reviewer decisions. Passing validators, website tests and addon import checks cannot substitute for this human evidence gate. For an approved future change, use existing schema fields: actual fully covered builds in `tested_builds`, scoped limitations and a reviewed report reference in notes, and the required version/changelog entry. Do not invent evidence fields or source IDs. A source registry entry needs a real, suitable source and its required HTTPS URL; a private capture or local path is not a public source URL. Retain genuine historical tested builds, but do not carry them forward as proof of a changed body or new build. If a metadata-only promotion increments the record version, explicitly map the tested version and unchanged body hash to the promoted version in the report. Behavior/requirement changes need appropriate retesting and versioning under the [macro standard](MACRO_STANDARD.md). After a separately approved canonical promotion, update the specific evidence-dependent test expectations and run the repository's release gates: `npm run build:data`, `npm run validate:data`, `npm run lint`, `npm run typecheck`, `npm test`, and `npm run build`. Do not bypass untested-content guards or rewrite historical frozen reports to make a new status pass. Relevant game-build, spell/talent, body, requirement, locale or binding changes trigger scoped regression review; an old pass is not automatically current. **Completing this document or filling a form does not itself change any status.**