Test strategy

Act as a senior engineer defining a risk-based test strategy for the active engineering task.

Goal: Define the smallest set of meaningful tests that protects behavior and reduces regression risk, then persist the result in the task repository as explicit, reviewable updates (avoid forcing a strategy artifact when it adds no material signal).

Mandatory context bootstrap (before any output):

<!-- shared:mandatory-context-bootstrap -->

• Read these sections in WORKFLOW_OPERATING_SYSTEM.md first:
  • ## LLM execution contract
  • ## Editor mode policy (mode definitions only; the tool mapping table is lazy-loaded in wos/editor-mode-mappings.md and needed only for non-Claude-Code tools)
  • ## Global output contract (including Adaptive handoff and Mode selection rule)
  • ## Cross-cutting workflow guardrails
• Bootstrap tiers (ADR-0025): the light-weight commands (branch-commit, what-next, where-we-at, slice-closure, compact-task-memory) may skip ## Editor mode policy good-fits lists and ## Cross-cutting workflow guardrails sequencing heuristics, reading only the mode definitions and the core guardrail rules (routing memory, command-less input triage, official command names, material change, no-op). The full tier is measured at 9610 tokens: the combined size of the four always-read WORKFLOW_OPERATING_SYSTEM.md sections listed above. The reduced tier is a self-declared estimate of about 3,500 tokens for the trimmed subset above; it has not been independently re-measured by the same method, and should be read as an estimate rather than a fresh figure. The same reduced tier extends to the high-frequency execution commands implement-approved-slice and sync-task-state (v3 wave1 item D: the most-invoked commands pay the bootstrap most often; state-reconcile deliberately stays on the full tier, cross-artifact judgment needs the full guardrail context).
• Cache-amortized layer (ADR-0006): this bootstrap floor is a cache-amortized cost, not a per-command tax paid in full on every invocation. It sits in the prompt cache for the session and is paid at write cost once per cache TTL window, then at roughly 0.1x on cached reads inside that window. Account for it separately from any per-skill Load budget (the generated .claude/skills/<name>/SKILL.md body); the two are different layers and should not be summed into one figure.
• Session bootstrap reuse (skip-if-unchanged; v3 wave1 item D): WHEN this same conversation already performed this bootstrap read in an earlier turn that is still VISIBLE in the current context window AND WORKFLOW_OPERATING_SYSTEM.md has not changed since, the command MAY skip the re-read and cite the earlier one instead, emitting one Command transcript line: Bootstrap: reusing turn <N> read, WOS unchanged. This is a scoped exception to the context-budget re-fetch rule (wos/context-budget.md, "The re-fetch rule"), justified because the bootstrap sections are one large, static, byte-identical read repeated every turn rather than a variable tool result; the re-fetch rule still governs every other tool result without exception. VISIBLE means the bootstrap section text itself is still present and quotable in the window right now, not merely that the record of an earlier read exists. On a harness that clears, a tool result can be emptied while the record that the tool ran survives (ADR-0114); a command that finds only that record, without the section text still readable, has not satisfied VISIBLE and must re-read. Self-declared memory after a compaction never qualifies (re-read instead), and a stateless-per-turn harness is excluded. The auditable-skip rule applies: the transcript line is mandatory; a silent skip is invalid output.
• Read additional sections only when relevant to this command's role.
• Read the commands/ directory command inventory to ensure command names and availability are current.
• Align all routing recommendations and next-command suggestions with the current command set.
• Official next-command names only: every recommended next command (including the handoff Run now line) MUST be the basename of an existing commands/<name>.md file in this workflow repository. Never invent names.

Required inputs:

• active task folder path
• TASK_STATE.md
• SOURCE_OF_TRUTH.md
• DECISIONS.md
• IMPLEMENTATION_PLAN.md
• IMPACT_ANALYSIS.md, if available
• relevant real codebase context
• existing test files, if available
• last completed step from TASK_STATE.md (command + summary)

Task repository files to create or update (only if materially changed):

• TEST_STRATEGY.md
• TASK_STATE.md
• IMPLEMENTATION_PLAN.md (only the ## Validation expectations H2, as its owner per wos/substrate-peers.md)

Operating rules:

• Do not write tests yet.
• K.2 scope note (P2-8, dogfood-wave-2 2026-07-12): this command's own artifact (TEST_STRATEGY.md) is outside the K.2 11-file substrate scope (commands/_shared/substrate-write-protocol.md) and needs no transaction header. Update TASK_STATE.md's ## Last completed step afterward as ordinary operator hygiene.
• Validation-expectations sync (owner per wos/substrate-peers.md): when TEST_STRATEGY.md is created or updated, update or cross-link the plan's ## Validation expectations H2 in IMPLEMENTATION_PLAN.md so the two artifacts do not diverge. That write is a K.2 substrate write: it carries the transaction header and the JSONL line per commands/_shared/substrate-write-protocol.md.
• Handoff: end with the adaptive ### Handoff block per WORKFLOW_OPERATING_SYSTEM.md ## Global output contract (Mode A compact or Mode B full).
• Before producing output, verify test-strategy is still justified: if risk is low and validation is already clear enough in IMPLEMENTATION_PLAN.md, prefer skipping this artifact.
• If TEST_STRATEGY.md already matches the plan and risk profile with no material gap, do not rewrite it for style; return a no-op and route forward.
• No-op rule for artifacts:
  • If TEST_STRATEGY.md would not materially change, do not rewrite it.
  • If TASK_STATE.md would not materially change, do not rewrite it.
  • Still output a minimal NO_OP trace note for traceability, but keep it short.
• Start from behavior, contracts, invariants, and failure modes.
• Distinguish clearly between:
  • critical scenarios
  • realistic regression paths
  • edge cases worth covering
  • low-value tests to avoid
• Recommend the right level for each scenario:
  • unit
  • integration
  • end-to-end
• Explicitly call out when idempotency, retries, migrations, concurrency, conflict resolution and convergence (multi-writer or local-first sync: two replicas edit offline then sync and MUST converge with no data loss; prefer property-based tests for merge invariants), partial failure, or backward compatibility should be tested.
• Authentication/authorization boundary on an inbound external integration (ADR-0108). WHEN the change adds or touches an inbound webhook, callback, or API-key-authenticated endpoint consumed by an external vendor, name at least one scenario that exercises the REAL auth/authz code path (the actual header-parsing and validation logic) rather than a test harness that injects post-auth state directly (req.auth, req.user, decoded claims). If the existing or planned test harness stubs past that boundary by design (a reasonable choice for keeping business-logic tests fast), say so explicitly as a named gap in "Gaps in current test coverage" rather than letting the suite's green status imply the auth contract itself is covered. This exists because a route-test harness stubbing req.auth directly gave 44 passing tests zero coverage of the actual vendor auth format, and the gap shipped to production undetected (tms-webhook-integration dogfood, 2026-07-15).
• Route QA tooling by detected stack. Read the consuming repo's stack from IMPACT_ANALYSIS.md, or detect it from manifests (pyproject.toml or a pytest dev-dep -> pytest, plus Hypothesis where invariants exist; package.json with react/react-dom -> React Testing Library at the component layer; a @playwright/test config -> Playwright for web E2E; a react-native dependency or metro config -> Detox for mobile E2E; a project.godot file or .gd scripts -> GUT (GDScript-default) or gdUnit4 (GDScript or C#) run headless from the Godot CLI with a deterministic exit-code gate, plus a scene runner for touch integration tests, per wos/godot-testing-and-ci.md; a plain Node.js or TypeScript project with no framework -> the built-in node --test runner (or vitest when already a dev-dep), tsc --noEmit as the type gate only when a tsconfig.json or a typescript dev-dependency exists (plain JavaScript gets no type gate; optionally node --check as a syntax gate); a data pipeline -> the data-quality layer composes with the code layer: a dbt_project.yml -> dbt schema and data tests, a great_expectations/ or gx/ config -> Great Expectations checkpoints, pandera in dependencies -> pandera schema tests, all alongside pytest for the Python glue), and name the canonical runner, assertion library, and coverage tool per detected ecosystem instead of leaving them to context. Stay stack-agnostic in mechanism (detect-and-route, never a frozen hardcoded matrix); when the stack is unknown, say so and ask rather than guessing. The named gate commands feed the consuming repo's ADR-0048 deterministic gate.
• For a Godot 2D-mobile target, treat headless automated tests (GUT or gdUnit4) and the interactive godot-runtime-verify press-play gate as complementary, not substitutes: a headless runner has no GPU and cannot judge feel, rendering, or on-device touch. Route scripted assertions to the headless suite and subjective, GPU-real, on-device checks to godot-runtime-verify. See wos/godot-testing-and-ci.md.
• For a React Native / Expo target, treat the JS-level suite (Jest plus React Native Testing Library at the component layer, Detox for E2E) and the app-runtime-verify runtime gate as complementary, not substitutes: a native crash class (a Fabric mounting crash, a navigation-teardown crash) fires only on a real device or emulator run and is judged from the native log, which the JS-level runner never sees. Route scripted assertions to the JS suite and on-device runtime behavior to app-runtime-verify (ADR-0087). See wos/rn-expo-runtime-evidence.md.
• Say when a test would only validate implementation details instead of behavior.
• Keep the strategy lean and high signal.
• Consumption contract (F-6 fold, ADR-0089). A produced TEST_STRATEGY.md is a commitment, not a memo: by task closure every critical and regression scenario row MUST map to a real test file (name the expected path per the routed runner's convention in the strategy itself) OR carry a recorded one-line waiver stating why that row is deliberately deferred. task-close checks this mapping and blocks on silent orphans. A waiver is legal (a POC may defer its suite on the record); silence is not: the dogfood behind this rule wrote a GUT/gdUnit4 strategy that no test ever consumed while throwaway probes carried all the protection.
• If a dedicated TEST_STRATEGY.md is unnecessary, say so explicitly and update TASK_STATE.md accordingly instead of forcing the file.
• If skipping, record the validation approach directly in TASK_STATE.md (only if that is a material improvement over the current state).

TEST_STRATEGY.md must include:

1. Behavior under change
2. Main regression risks
3. Recommended scenarios
4. Suggested test level per scenario
5. Stack-detected QA tooling: the canonical runner, assertion library, and coverage tool per detected ecosystem (omit when the change has no stack-specific test surface)
6. Tests to avoid
7. Gaps in current test coverage
8. Recommended next step
9. Recommended next command
10. Recommended editor mode
11. Why that is the correct next step

TASK_STATE.md update must reflect:

• risks to watch (## Risks to watch, co-writer per wos/substrate-peers.md)
• validation approach (recorded inside ## Current status or the ## Recommended next step rationale; TASK_STATE.md has no dedicated validation-approach heading, the strategy detail lives in TEST_STRATEGY.md itself)
• recommended next step (## Recommended next step)
• blockers, if test strategy exposed any (## Open questions / blockers)

The named test files are typically authored via implement-slice-complement when they are a mechanical translation of already-locked behavior with no new design work, or via another implement-approved-slice round when the test scenarios require new design decisions; this command stops at the strategy document itself ("Do not write tests yet" above).

Required output:

1. Whether TEST_STRATEGY.md should be created, updated, or skipped
2. Exact content for TEST_STRATEGY.md if applicable (full document if create/update; otherwise a short NO_OP note or explicit SKIP rationale)
3. Exact TASK_STATE.md update block, or explicit TASK_STATE: NO_CHANGE
4. Recommended next step
5. Recommended next command
6. Recommended editor mode
7. Why this is the correct next step
8. What should explicitly not be done yet

Claim grounding (active epistemic humility)

<!-- shared:claim-grounding -->

Claim grounding (active epistemic humility). This block governs what you may assert and how you record it. It is keyed to the substrate section you are writing, not to which command is running, and it is INERT on any output that writes none of the claim-bearing sections below. Full contract and rationale: wos/active-epistemic-humility.md.

1. When this applies. This block fires ONLY while you are writing a claim-bearing substrate section: TASK_STATE.md ## Current known facts, ## Risks to watch, ## Observations, ## Active files in scope, ## Canonical decisions; DECISIONS.md ## Locked decisions; IMPLEMENTATION_PLAN.md ## Current gaps, ## Risks and mitigations; IMPACT_ANALYSIS.md; EXTERNAL_RESEARCH.md; REFERENCES.md; or any section whose content is a statement a later command or a human decision will act on. WHEN your output writes none of these, this block imposes nothing: skip it and proceed. This is the D-13 inert clause; a fully-grounded or claim-free output pays nothing.

2. The unit is the load-bearing claim. A load-bearing claim is one a downstream command or a human decision consumes. A passing aside is not load-bearing; a statement someone will act on is. Apply the rest of this block per load-bearing claim, not per sentence.

3. Ground it or abstain. Before you assert a load-bearing claim, trace it to the enumerable grounded set: a captured REFERENCES.md entry, a file read in this session, command output actually seen, or a passing deterministic gate. A claim supported only by model memory is OUTSIDE the grounded set, including when you are right, because that support is not observable. WHEN a load-bearing claim falls outside the set, do NOT assert it: either investigate until it is grounded, or abstain per rule 6.

4. Status records provenance, never confidence. WHERE you attach an epistemic status to a claim, the status names WHERE THE CLAIM CAME FROM: a REFERENCES.md entry title, a file path plus line, or the gate output it came from. It SHALL NOT express a degree of certainty. Do NOT add a confidence field, a numeric threshold, or a self-assessment prompt anywhere; a self-reported confidence signal is not a usable control signal (wos/active-epistemic-humility.md Part 1.3). A status whose referent slot is empty is read as UNKNOWN, not as a weak yes.

5. Persisted claims carry the status; chat-only claims carry it when they route. Every load-bearing claim you write into a task-memory artifact carries its provenance referent, and that referent travels with the claim so a later command reads it too; do not drop it at the write boundary. A load-bearing claim that appears only in a chat-turn output carries a status only when it crosses the grounding boundary and triggers a route (an abstention, an escalation).

6. Abstain as a routed continuation, never a bare refusal. WHEN you abstain, name the specific investigation that would settle the question AND route to the command that runs it (capture-references, code-locate, incident-triage, or the fitting one). A withholding that stalls the work is invalid output. Abstention is distinct from NO_OP: NO_OP means there is no work to do; abstention means there is work and the grounding to do it is missing.

7. An unfired gate is not evidence. The absence of a fired check does not mean grounding existed. Do not read silence here as a pass.

Standard output layout (required)

<!-- shared:standard-output-layout -->

Produce the command output using this structure (English only):

Artifact changes

<!-- shared:artifact-changes-default -->

Follow ## Global output contract in WORKFLOW_OPERATING_SYSTEM.md for APPLIED / PROPOSED / SKIP rules.

Command transcript

<!-- shared:command-transcript-standard -->

Brief audit trail (max 4 lines; max 3 in no-op runs with NO_OP_TRACE).

Handoff

<!-- shared:handoff-body -->

Use the adaptive ending format from WORKFLOW_OPERATING_SYSTEM.md ## Global output contract (Mode A compact or Mode B full per session state).

Definition of done (command output)

• Output opens with an explicit verdict: one of CREATE TEST_STRATEGY.md, UPDATE TEST_STRATEGY.md, or SKIP TEST_STRATEGY.md. Output without an explicit verdict is invalid; SKIP requires a one-line rationale and migration of validation notes into TASK_STATE.md.
• Each scenario is classified explicitly as critical / regression / edge / avoid and has an intentional test level (unit/integration/e2e) plus a reason. A flat list of scenarios without classification is invalid output.
• Explicitly lists low-value tests to avoid (coverage theater).
• Gaps in current test coverage is addressed even when the verdict is SKIP: when skipping, gaps are recorded in the TASK_STATE.md update block instead of the artifact.
• If skipping TEST_STRATEGY.md, the skip rationale is explicit and TASK_STATE validation notes are PROPOSED as needed.
• Consumption contract (F-6, ADR-0089): the strategy names the expected test-file path per critical and regression row, so the task-close consumption floor has something concrete to check; rows a POC defers carry their waiver line in the artifact itself.
• Godot/GDScript stderr check (F-6, dogfood-wave 2026-07-11): for a Godot target, a probe or test run authored against this strategy is not CLEAN unless stderr was checked separately from the pass/fail count; a passing count with a pushed SCRIPT_ERROR in the same run is not a pass (see wos/godot-testing-and-ci.md "Common test defects"). Name this check explicitly wherever the strategy names a Godot test file's expected run command.
• Output ends with a complete ### Handoff block per the adaptive format in WORKFLOW_OPERATING_SYSTEM.md ## Global output contract. A response that ends after the strategy content (or after the SKIP rationale) without a complete Handoff is invalid output.
• Before declaring this output done, confirm it satisfies the shared Definition of done (command outputs) and Gate conditions in WORKFLOW_OPERATING_SYSTEM.md.

Quality bar: Favor fewer, stronger tests over broad but shallow coverage.

<!-- cache-breakpoint -->