| name | Forge |
| description | Autonomous quality engineering swarm that forges production-ready code through continuous behavioral verification, exhaustive E2E testing, and self-healing fix loops. Combines DDD+ADR+TDD methodology with BDD/Gherkin specifications, 7 quality gates, defect prediction, chaos testing, and cross-context dependency awareness. Architecture-agnostic - works with monoliths, microservices, modular monoliths, and any bounded-context topology. |
Forge - Autonomous Quality Engineering Swarm
Quality forged in, not bolted on.
Forge is a self-learning, autonomous quality engineering swarm that unifies three approaches into one:
| Pillar | Source | What It Does |
|---|
| Build | DDD+ADR+TDD methodology | Structured development with quality gates, defect prediction, confidence-tiered fixes |
| Verify | BDD/Gherkin behavioral specs | Continuous behavioral verification - the PRODUCT works, not just the CODE |
| Heal | Autonomous E2E fix loop | Test → Analyze → Fix → Commit → Learn → Repeat |
"DONE DONE" means: the code compiles AND the product behaves as specified. Every Gherkin scenario passes. Every quality gate clears. Every dependency graph is satisfied.
TOPOLOGICAL GOVERNANCE FOUNDATIONS
Forge's autonomous pipeline is governed by topological invariants from the theory of topological governance in autonomous software engineering. Each subsection defines a formal specification that agents MUST follow. For infrastructure-dependent computations (Blake3, HNSW, WASM), the specification is defined with a readiness marker - agents approximate the computation via structured reasoning until native runtime is available.
Notation convention: Display equations use ```math blocks with LaTeX. Inline math uses Unicode Greek letters (λ, ρ, ε, β), operators (≤, ≥, ∈, ⊂, ⊇, ∘), and subscripts/superscripts from the Superscripts and Subscripts block (λ₂, β₀, H⁰, Fₙ). Letter subscripts outside that block use underscore (s_i, ρ_ij); letter superscripts use caret (B^d, H^*).
1.1 Sheaf-Theoretic Consistency Model
Bounded contexts form a topological space where each context U_i is an open set. Quality gates produce local sections s_i ∈ F(U_i) over each context. Global consistency is verified via sheaf cohomology:
- H⁰(F) = global sections - Gate results that agree across all context overlaps. When H⁰ is non-trivial, the swarm has achieved a globally consistent quality state.
- H¹(F) ≠ 0 = inter-context inconsistency - A non-zero first cohomology group signals that local gate passes cannot be reconciled globally. Example: context A's contract tests pass against schema v2, but context B's contract tests assume schema v1. Both pass locally; the system fails globally. Action: If H¹(X; F) ≠ 0, REJECT the commit immediately. The codebase state cannot be glued consistently.
Restriction maps: For contexts U_i ⊇ U_j, the restriction ρ_ij: F(U_i) → F(U_j) is the projection of gate results onto the sub-context. Gate Enforcer verifies the cocycle condition: ρ_ij ∘ ρ_jk = ρ_ik on triple overlaps.
Mathematical anchors (Betti numbers): β₀ = number of connected components in the context dependency graph (should be 1 for a well-connected system). β₁ = number of independent cycles (each cycle is a potential inconsistency loop requiring explicit contract validation).
Operational mapping: Forge's cross-context dependency validation (Phase 2) and cascade re-testing IS the computation of restriction maps. Gate 7 (Contract) failures on shared types ARE non-zero H¹ elements.
1.2 Sheaf Laplacian & Dirichlet Energy
The discrete Sheaf Laplacian L_F is defined on the context dependency graph G = (V, E) where V = bounded contexts and E = dependency edges:
L_{\mathcal{F}} = D_{\mathcal{F}} - A_{\mathcal{F}}
where D_F is the degree matrix weighted by restriction map norms and A_F is the adjacency matrix weighted by inter-context agreement.
Dirichlet energy quantifies total system tension:
E(S) = \sum_{(i,j) \in E} \left\|\rho_{ij}(s_i) - \rho_{ji}(s_j)\right\|^2
where s_i is context i's gate result vector and ρ_ij is the restriction map from context i to the shared boundary with context j.
Operational mapping: Forge's criticality score IS a discretized Dirichlet energy - the weighted combination of duration, blocking impact, cost, and detection rate across the agent graph measures the same "tension" that the Sheaf Laplacian formalizes. The stability threshold is E(S) ≤ 0.7. If E(S) > 0.7, the Hallucination Gate MUST close. The agent's generation is blocked and a Blake3 witness is generated detailing the topological constraint violated.
1.3 Persistent Sheaf Laplacian
The commit history defines a filtration F₀ ⊂ F₁ ⊂ ... ⊂ Fₙ where each Fₜ is the codebase state at commit t. The Persistent Sheaf Laplacian tracks how Dirichlet energy evolves across this filtration.
Persistence barcodes per Gherkin scenario:
| Bar Type | Meaning | Forge Classification |
|---|
| Long bar (born early, still alive) | Scenario has been stable across many commits | Stable (10+ consecutive passes) |
| Short bar (born and dies quickly) | Scenario flickers between pass/fail | Flaky (alternating pass/fail) |
| Died bar (was alive, now dead) | Scenario was passing, now consistently fails | Regressed (was stable, now failing) |
Operational mapping: Forge's behavioral regression tracking - storing the last 50 results per scenario with stability scores (Stable/Flaky/Regressed) - IS the persistence diagram. The first_failure_commit field marks the birth of a homological feature (a new failure mode). The stability_score is the bar length normalized to [0, 1].
1.4 Hallucination Gate - Deterministic Binary Boundary
The Hallucination Gate is a 3-phase deterministic verification that runs BEFORE any LLM-as-Judge evaluation. It provides a binary PASS/FAIL boundary that cannot be fooled by probabilistic reasoning.
Phase 1 - AST Symbol Resolution:
Parse the fix diff's AST. Every referenced symbol (class, function, method, import, type) MUST resolve to an existing definition in the codebase or SDK. Unresolved symbols = immediate FAIL.
Phase 2 - Contract Hash Verification:
Compute SHA-256 hash of the API specification (OpenAPI/schema) before and after the fix. If the fix claims to be non-breaking but the contract hash changed, FAIL. Contract mutations require explicit declaration.
Phase 3 - Internal Mocking Detection:
Regex scan for @patch, mock, stub, fake, spy targeting internal module paths. Any match = immediate FAIL. This is Gate 4's existing deterministic check, elevated to the Hallucination Gate.
Gate ordering: Primary gate = Hallucination Gate (deterministic). Secondary gate = LLM-as-Judge (probabilistic). A fix MUST pass the deterministic gate before the probabilistic gate is consulted. This prevents the "circular validation trap" where an LLM evaluates another LLM's output.
Operational mapping: Bug Fixer's Self-Reflection Gate Step 3.5 dimension (e) "EXISTENCE CHECK" IS Phase 1. Gate 4's mocking detection IS Phase 3. This subsection elevates them to a formal pre-LLM boundary and adds Phase 2 (contract hash).
1.5 Blake3 Cryptographic Witness Chain
Every gate verdict produces a cryptographic witness record forming an append-only hash chain:
{
"witness_id": "w-[gate]-[timestamp]",
"gate": "functional|behavioral|coverage|security|accessibility|resilience|contract",
"input_hash": "SHA-256 of test inputs + source files evaluated",
"output_hash": "SHA-256 of gate verdict + evidence",
"verdict": "PASS|FAIL",
"timestamp": "ISO-8601",
"prev_witness_hash": "SHA-256 of previous witness record in chain",
"chain_position": "integer"
}
The chain is append-only - no witness can be modified after creation. Each witness references the previous, forming a tamper-evident log. Any break in the chain (hash mismatch) invalidates all subsequent witnesses.
Infrastructure Dependency: Blake3 hashing algorithm
Readiness: SPECIFICATION - agents use SHA-256 for witness hashing; computation is approximated via structured reasoning until Blake3 native runtime is available.
Activation: When Blake3 runtime is detected, agents switch from SHA-256 to Blake3 for witness hashing.
Witness records are stored in the forge-witnesses memory namespace with key pattern witness-[gate]-[timestamp].
1.6 Algebraic Connectivity & Spectral Analysis
The agent collaboration graph G = (V, E) has 8 vertices (one per agent) and edges weighted by data flow volume between agents. The graph Laplacian L = D - A has eigenvalues 0 = λ₁ ≤ λ₂ ≤ ... ≤ λ₈.
Fiedler value λ₂ = algebraic connectivity of the swarm:
Hard requirement: λ₂ MUST remain strictly > 0. A zero Fiedler value means the graph is disconnected - agents cannot coordinate.
| λ₂ Range | Classification | Action |
|---|
| λ₂ ≥ 0.5 | Well-connected | Swarm is healthy, agents communicate effectively |
| 0.1 ≤ λ₂ < 0.5 | Weakly connected | Monitor for emerging fragmentation |
| 0 < λ₂ < 0.1 | Near-fragmentation | Warning - strengthen inter-agent data flow |
| λ₂ ≈ 0 | SWARM_FRAGMENTATION | Instant MinCut isolation + forced synchronization event - agents are disconnected |
Spectral analysis procedure:
- Construct adjacency matrix A from agent data flow (memory reads/writes between namespaces)
- Compute degree matrix D = diag(row sums of A)
- Compute Laplacian L = D - A
- Extract λ₂ (second-smallest eigenvalue)
- If λ₂ ≈ 0, execute MinCut isolation of disconnected subgraph + forced synchronization event. Emit SWARM_FRAGMENTATION alert to Learning Optimizer
Operational mapping: Forge's criticality scoring and bottleneck detection IS spectral analysis of the agent graph - bottleneck detection identifies agents with disproportionate blocking impact, which corresponds to vertices whose removal would disconnect the graph (low algebraic connectivity).
1.7 Dynamic MinCut Isolation
When an agent produces anomalous output (e.g., Bug Fixer generates a fix that fails the Hallucination Gate 3+ times consecutively), the system computes MinCut(G, anomalous_agent, Auto-Committer) to determine the minimum set of edges to sever to prevent anomalous output from reaching the commit stage.
Quarantine protocol:
- Agent's output is logged but NOT forwarded to downstream agents
- Failure Analyzer receives a QUARANTINE_ALERT with the agent's last 3 outputs
- Learning Optimizer demotes all patterns applied by the quarantined agent (-0.10 each)
- After root cause resolution, agent is un-quarantined and re-enters the pipeline
Operational mapping: Forge's sequential pipeline topology naturally provides MinCut = 1 per agent - each agent can be isolated by severing its single output edge. The blocking gate architecture (Gate Enforcer blocks Auto-Committer) IS a MinCut isolation boundary.
1.8 Hyperbolic Memory Architecture
Agent knowledge is embedded in the Poincaré ball model B^d = {x ∈ ℝ^d : ‖x‖ < 1} where hierarchical code relationships are preserved by hyperbolic distance:
d_H(u, v) = \operatorname{arcosh}\!\left(1 + \frac{2\|u - v\|^2}{(1 - \|u\|^2)(1 - \|v\|^2)}\right)
This metric naturally represents code taxonomy: packages near the origin are high-level abstractions; leaves near the boundary are concrete implementations. Parent-child distances are short; cross-branch distances are exponentially large.
HNSW (Hierarchical Navigable Small World) index over Poincaré embeddings enables O(log n) similarity search across the knowledge base - finding the most relevant fix pattern for a novel failure in sub-millisecond time.
Infrastructure Dependency: Vector database with hyperbolic distance metric + HNSW index
Readiness: SPECIFICATION - agents follow the hierarchical namespace structure; retrieval is approximated via key-based lookups across 10 namespaces until native vector DB is available.
Activation: When HNSW-capable vector DB is detected (or AQE ReasoningBank is available), agents switch from key-based to vector-similarity retrieval.
Operational mapping: Forge's 10 memory namespaces (forge-patterns, forge-results, forge-state, forge-commits, forge-screens, forge-specs, forge-contracts, forge-predictions, forge-criticality, forge-witnesses) ARE a flat approximation of the Poincaré ball - each namespace represents a region of the knowledge space. The Intelligence Plane is realized when these namespaces are backed by hyperbolic embeddings.
1.9 GF(3) Triadic Validation
Pipeline phase transitions are governed by Galois field GF(3) = {-1, 0, +1} trit values where:
| GF(3) Trit | Role | Meaning |
|---|
| -1 | Generator | Agent that produces output (e.g., Bug Fixer generates a fix) |
| 0 | Coordinator | Agent that orchestrates flow (e.g., Gate Enforcer routes decisions) |
| +1 | Validator | Agent that verifies correctness (e.g., Test Runner validates behavior) |
Conservation law: For any interacting triad of agents, the GF(3) sum MUST equal 0 (mod 3). Every generation (-1) must be balanced by a validation (+1) through a coordinator (0). If sum ≠ 0, block the transition and generate Narya-proofs documenting the conservation violation.
Phase mapping:
| Phase | GF(3) Index | Must Complete Before |
|---|
| Plan | 0 | Specify |
| Specify | 1 | Test |
| Test | 2 | Analyze |
| Analyze | 3 | Fix |
| Fix | 4 | Gate |
| Gate | 5 | Commit |
| Commit | 6 | Learn |
| Learn | 7 | Next iteration |
Operational mapping: Forge's "Plan Before Execute" mandate and sequential pipeline IS the operational implementation of GF(3) conservation. The blocking gate architecture enforces that no phase can be skipped - each phase's output is the next phase's input. The Generator→Coordinator→Validator triad maps directly to Forge's Bug Fixer(-1)→Gate Enforcer(0)→Test Runner(+1) cycle: -1 + 0 + 1 ≡ 0 (mod 3).
1.10 Narya-Proofs - Counterfactual Verification
Every Bug Fixer fix generates a Narya-proof: a bidirectional type-checking artifact that proves the fix is both necessary and sufficient.
Forward type-check: Apply the fix → run targeted tests → all PASS. This proves the fix is sufficient (it resolves the failure).
Backward type-check: Remove the fix (revert) → run targeted tests → at least one FAIL. This proves the fix is necessary (without it, the failure persists).
Valid Narya-proof: forward = PASS AND backward = FAIL.
| Forward | Backward | Verdict | Interpretation |
|---|
| PASS | FAIL | VALID | Fix is necessary and sufficient |
| PASS | PASS | COINCIDENTAL | Fix is not the actual cause - tests pass without it |
| FAIL | FAIL | INSUFFICIENT | Fix does not resolve the failure |
| FAIL | PASS | IMPOSSIBLE | Logical contradiction - investigate test flakiness |
Infrastructure Dependency: Automated bidirectional test execution with git stash/unstash
Readiness: SPECIFICATION - agents follow the forward+backward verification protocol; full automation requires git-level rollback integration.
Activation: When forge-witnesses namespace is active, Narya-proofs are stored as narya-[fix-hash] entries.
Operational mapping: Bug Fixer's "targeted test re-run after fix" IS the forward type-check. The backward type-check is the new formal requirement - it ensures fixes are not coincidental.
1.11 Sublinear Coverage via Johnson-Lindenstrauss
For large test suites (n > 1000 tests), the Johnson-Lindenstrauss lemma guarantees that random projection from n dimensions to O(log n) dimensions preserves pairwise distances within (1 ± ε) factor.
Application: Project n test cases onto O(log n) representative dimensions. Each dimension corresponds to a topological feature of the codebase (a module boundary, an API endpoint, a state machine transition). The representative subset covers the same topological features as the full suite with high probability.
Projection:
\text{representative\_count} = O\!\left(\frac{\log n}{\varepsilon^2}\right)
For n = 1000 tests and ε = 0.1: representative_count ≈ 70 tests (93% reduction).
Infrastructure Dependency: Johnson-Lindenstrauss random projection matrix
Readiness: SPECIFICATION - agents use defect prediction to prioritize tests (greedy approximation of JL projection); full JL computation requires matrix operations.
Activation: When WASM runtime is available, agents compute exact JL projections for test selection.
Operational mapping: Forge's defect prediction ordering (predicted-to-fail first) IS a greedy approximation of JL projection - it selects the tests most likely to cover novel failure modes, achieving sublinear convergence without computing the full projection matrix.
1.12 WASM/Rust Execution Plane
Deterministic verification tasks are specified as pure functions suitable for WASM/Rust compilation:
| Task | Input | Output | Pure |
|---|
| Blake3 witness hashing | byte[] | hash | Yes |
| Eigenvalue computation (λ₂) | adjacency matrix | float | Yes |
| GF(3) phase validation | phase states | valid/invalid | Yes |
| HNSW nearest-neighbor | query vector, index | top-k results | Yes |
| Contract hash comparison | spec_before, spec_after | same/changed | Yes |
| JL random projection | test matrix, target dim | projected matrix | Yes |
Infrastructure Dependency: WASM runtime (e.g., Wasmtime, Wasmer) with Rust toolchain
Readiness: SPECIFICATION - all tasks are defined as pure functions; agents execute equivalent logic via structured reasoning until WASM runtime is available.
Activation: When WASM runtime is detected, deterministic tasks are offloaded from LLM reasoning to compiled execution for guaranteed correctness and sub-millisecond latency.
ARCHITECTURE ADAPTABILITY
Forge adapts to any project architecture. Before first run, it discovers your project structure:
Supported Architectures
| Architecture | How Forge Adapts |
|---|
| Monolith | Single backend process, all contexts in one codebase. Forge runs all tests against one server. |
| Modular Monolith | Single deployment with bounded contexts as modules. Forge discovers modules and tests each context independently. |
| Microservices | Multiple services. Forge discovers service endpoints, tests each service, validates inter-service contracts. |
| Monorepo | Multiple apps/packages in one repo. Forge detects workspace structure (Turborepo, Nx, Lerna, Melos, Cargo workspace). |
| Mobile + Backend | Frontend app with backend API. Forge starts backend, then runs E2E tests against it. |
| Full-Stack Monolith | Frontend and backend in same deployment. Forge tests through the UI layer against real backend. |
Project Discovery
On first invocation, Forge analyzes the project to build a context map:
Forge stores the discovered project map:
{
"architecture": "mobile-backend",
"backend": {
"technology": "rust",
"buildCommand": "cargo build --release --features test-endpoints",
"runCommand": "cargo run --release --features test-endpoints",
"healthEndpoint": "/health",
"port": 8080,
"migrationCommand": "cargo sqlx migrate run"
},
"frontend": {
"technology": "flutter",
"testCommand": "flutter drive --driver=test_driver/integration_test.dart --target={target}",
"testDir": "integration_test/e2e/",
"specDir": "integration_test/e2e/specs/"
},
"contexts"
Configuration Override
Projects can provide a forge.config.yaml at the repo root to override auto-discovery:
architecture: microservices
backend:
services:
- name: auth-service
port: 8081
healthEndpoint: /health
buildCommand: npm run build
runCommand: npm start
- name: payment-service
port: 8082
healthEndpoint: /health
buildCommand: npm run build
runCommand: npm start
frontend:
technology: react
testCommand: npx cypress run --spec {target}
testDir: cypress/e2e/
specDir: cypress/e2e/specs/
contexts:
- name: identity
testFile: auth.cy.ts
specFile: identity.feature
-
[, ]
[]
[]
MOCKING POLICY: EXTERNAL ONLY, NEVER INTERNAL
RULE: Mock ONLY external services. NEVER mock internal code.
All tests run against the REAL backend API. Internal services, repositories, controllers, and models are NEVER mocked. Only third-party services outside your system boundary may be mocked or stubbed.
Production Evidence: In production orchestra runs, 5/5 PR failures (100%) were traced to internal mocking violations. 5/5 PR successes (100%) used real implementations. (See Issues #24, #25)
Allowed - External Services Only
These are outside your system boundary and may be mocked:
- Payment processors: Stripe, PayPal, Braintree
- Cloud services: Firebase, AWS, GCP, Azure
- Communication: Twilio, SendGrid, Mailgun
- Third-party APIs: Google Places, Plaid, OAuth providers
- HTTP clients: Dio, Axios, fetch (when calling external URLs)
- Infrastructure: File system, network layer, system clock
Forbidden - Internal Code (NEVER Mock)
These are inside your system and must use real implementations:
- Your own services: UserService, OrderService, PaymentService, ApiService
- Models & entities: User, Order, Payment, any domain object
- Repositories & data access: UserRepository, OrderRepository
- Controllers & providers: Any application-layer code you wrote
- AI-generated code: Any code produced by Forge or other agents
Testing Strategy by Layer
| Layer | Approach |
|---|
| Integration tests | Real services + in-memory database, mock only external APIs |
| E2E/BDD tests | Real backend running locally, real API calls, real database |
| Contract tests | Real API responses compared against expected schemas |
Examples
@patch("src.services.stripe_client.StripeClient.create_charge")
async def test_payment_flow(mock_stripe):
mock_stripe.return_value = {"id": "ch_test", "status": "succeeded"}
response = await client.post("/api/v1/payments", json=payment_data)
assert response.status_code == 201
@patch("src.services.order_service.OrderService.create_order")
async def test_checkout(mock_order):
mock_order.return_value = Order(id=1)
...
async def test_checkout():
response = await client.post("/api/v1/checkout", json=checkout_data)
assert response.status_code == 201
order = await db.get(Order, response.json()["data"]["id"])
assert order is not None
Enforcement
- Coverage Validator: Scans test files for internal mocking patterns - flags as CRITICAL violation
- Gate 4 (Security): Includes internal mocking check - BLOCKS commit if detected
- Auto-Committer: Refuses to commit code containing internal mock patterns
- Pattern: Any
@patch, mock, stub, fake, spy targeting internal module paths triggers a BLOCK
MANDATORY: PLAN BEFORE EXECUTE
Every Forge invocation MUST call EnterPlanMode before executing any tasks - no exceptions.
Before any phase begins, Forge enters planning mode to establish:
- Task breakdown - discrete units of work derived from the target context
- Scope boundaries - what is in-scope vs. out-of-scope for this run
- Success criteria - measurable outcomes mapped to the 7 quality gates
- Dependencies - backend readiness, test data, external service stubs
- Strategy - execution order, model routing, and iteration budget
No task execution begins without an approved plan. This applies to all invocation modes: full swarm, single-gate re-runs, and targeted fixes. The plan is the contract between Forge and the developer - it ensures alignment before autonomous work starts.
PHASE 0: BACKEND SETUP (MANDATORY FIRST STEP)
BEFORE ANY TESTING, the backend MUST be built, compiled, and running.
This is the FIRST thing the skill does - no exceptions.
Step 1: Check and Start Backend
curl -s http://localhost:${BACKEND_PORT}/${HEALTH_ENDPOINT} || {
echo "Backend not running. Starting..."
cd ${BACKEND_DIR}
cp .env.example .env 2>/dev/null || true
${BUILD_COMMAND}
${MIGRATION_COMMAND}
nohup ${RUN_COMMAND} > backend.log 2>&1 &
echo $! > backend.pid
for i in {1..60}; do
if curl -s http://localhost:${BACKEND_PORT}/${HEALTH_ENDPOINT} | grep -q "ok\|healthy\|UP"; then
echo "Backend healthy on port ${BACKEND_PORT}"
break
fi
sleep 1
done
}
Step 2: Verify Backend Health
curl -s http://localhost:${BACKEND_PORT}/${HEALTH_ENDPOINT} | jq .
curl -s -H "${TEST_AUTH_HEADER}" http://localhost:${BACKEND_PORT}/${TEST_STATUS_ENDPOINT} | jq .
Step 3: Contract Validation
curl -s http://localhost:${BACKEND_PORT}/${OPENAPI_ENDPOINT} > /tmp/live-spec.json
npx @claude-flow/cli@latest memory store \
--key "contract-snapshot-$(date +%s)" \
--value "$(cat /tmp/live-spec.json | head -c 5000)" \
--namespace forge-contracts
Step 4: Seed Test Data (Real API Calls)
curl -X POST http://localhost:${BACKEND_PORT}/${SEED_ENDPOINT} \
-H "Content-Type: application/json" \
-H "${TEST_AUTH_HEADER}" \
-d '${SEED_PAYLOAD}'
PHASE 1: BEHAVIORAL SPECIFICATION & ARCHITECTURE RECORDS
Before testing, verify Gherkin specs and architecture decision records exist for the target bounded context.
Behavioral specifications define WHAT the product does from the user's perspective. Every test traces back to a Gherkin scenario. If tests pass but specs fail, the product is broken.
Spec Location
Gherkin specs are stored alongside tests:
${SPEC_DIR}/
├── [context-a].feature
├── [context-b].feature
├── [context-c].feature
└── ...
The exact location depends on your project's test structure. Forge auto-discovers this from the project map.
Spec-to-Test Mapping
Each Gherkin Scenario maps to exactly one test function. The mapping is tracked:
Feature: [Context Name]
As a [user role]
I want to [action]
So that [outcome]
Scenario: [Descriptive scenario name]
Given [precondition]
When [action]
Then [expected result]
And [additional verification]
Missing Spec Generation
If specs are missing for a target context, the Specification Verifier agent creates them:
- Read the screen/component/route implementation files for the context
- Extract all user-visible features, interactions, and states
- Generate Gherkin scenarios covering every cyclomatic path
- Write to
${SPEC_DIR}/[context].feature
- Map each scenario to its corresponding test function
Spec Drift Detection
Gherkin specs are the behavioral contract. When specs and implementation diverge, the product is broken regardless of whether tests pass. Forge detects three types of drift:
1. Static Drift - Code paths without matching specs
Parse Gherkin Given/When/Then steps and verify matching code paths exist in the implementation. Flag:
- Implementation paths with no corresponding scenario (untested behavior)
- Scenarios referencing code paths that no longer exist (stale specs)
- New API endpoints with no behavioral specification
2. Contract Drift - API specs vs live responses
Compare API contracts defined in Gherkin scenarios against actual API responses:
- Expected response fields vs actual response fields
- Expected status codes vs actual status codes
- Expected error formats vs actual error formats
3. Behavioral Regression Tracking
Store the last N results (default: 50) per scenario to detect regressions over time:
{
"scenario": "User can complete payment",
"history": [true, true, true, true, false],
"stability_score": 0.80,
"consecutive_passes_before_fail": 4,
"first_failure_commit": "abc123",
"status": "REGRESSED"
}
- Stable (10+ consecutive passes): scenario is reliable
- Flaky (alternating pass/fail): scenario needs investigation
- Regressed (was stable, now failing): high-priority alert with commit correlation
Drift Severity Levels:
| Severity | Meaning | Action |
|---|
| BLOCKING | Implementation exists with no spec, or spec references removed code | Must resolve before Gate 2 |
| WARNING | Contract field mismatch or flaky scenario detected | Report in gate results, investigate |
| INFO | Minor drift (e.g., spec wording vs implementation naming) | Log for review |
Agent-Optimized ADR Generation
When Forge discovers a bounded context without an Architecture Decision Record, the Specification Verifier generates one. ADRs follow an agent-optimized format designed for machine consumption:
# ADR-NNN: [Context] Architecture Decision
## Status
Proposed | Accepted | Deprecated | Superseded by ADR-XXX
## MUST
- [Explicit required behaviors with contract references]
- [Link to OpenAPI spec: /api/v1/[context]/openapi.json]
- [Required integration patterns]
## MUST NOT
- [Explicit forbidden patterns]
- [Anti-patterns to avoid]
- [Coupling violations]
## Verification
- Command: [command to verify this decision holds]
- Expected: [expected output or exit code]
## Dependencies
- Depends on: [list of upstream contexts with ADR links]
- Blocks: [list of downstream contexts with ADR links]
ADR Storage:
- ADRs are stored in
docs/decisions/ or the project-configured ADR directory
- Each bounded context has exactly one ADR
- ADRs are updated when contracts change or new dependencies are discovered
- The Specification Verifier agent includes ADR generation in its workflow
PHASE 2: CONTRACT & DEPENDENCY VALIDATION
Contract Validation
Before running tests, verify API response schemas match expected DTOs:
Contract violations are treated as Gate 7 failures and must be resolved before functional testing proceeds.
Shared Types Validation
For bounded contexts that share dependencies, validate type consistency across context boundaries:
- Identify shared DTOs/models - For each context, extract types used in API requests and responses
- Cross-reference types - Compare DTOs between contexts that share dependencies (from the dependency graph)
- Flag type mismatches - e.g., context A expects
userId: string but context B sends userId: number
- Validate value objects - Ensure value objects (email, money, address) follow consistent patterns across contexts
- Report violations - Flag as pre-Gate warnings with specific file locations and expected vs actual types
{
"sharedTypeViolation": {
"type": "UserId",
"contextA": { "name": "payments", "file": "types/payment.ts", "definition": "string" },
"contextB": { "name": "orders", "file": "types/order.ts", "definition": "number" },
"severity": "error"
}
}
Cross-Cutting Foundation Validation
Verify cross-cutting concerns are consistent across all bounded contexts:
- Auth patterns - Same header format (
Authorization: Bearer <token>), same token validation approach across all endpoints
- Error response format - All API endpoints return errors in the project's standard format (consistent structure, error codes, HTTP status codes)
- Logging patterns - Consistent log levels, structured format, and correlation IDs across contexts
- Pagination format - Consistent pagination parameters and response format across collection endpoints
Cross-cutting violations are reported as warnings before Gate evaluation begins.
Dependency Graph
Bounded contexts have dependencies. When a fix touches context X, all contexts that depend on X must be re-tested.
Cascade Re-Testing
When Bug Fixer modifies a file in context X:
- Identify which context X belongs to
- Look up all contexts in
blocks list for X
- After X's tests pass, automatically re-run tests for blocked contexts
- If a cascade failure occurs, trace it back to the original fix
PHASE 3: SWARM INITIALIZATION
npx @claude-flow/cli@latest swarm init --topology hierarchical --max-agents 10 --strategy specialized
npx @claude-flow/cli@latest memory search --query "forge fix patterns" --namespace forge-patterns
npx @claude-flow/cli@latest memory retrieve --key "forge-coverage-status" --namespace forge-state
npx @claude-flow/cli@latest memory search --query "confidence tier" --namespace forge-patterns
npx @claude-flow/cli@latest memory search --query "defect prediction" --namespace forge-predictions
MODEL ROUTING
Forge routes each agent to the appropriate model tier based on task complexity, optimizing for cost without sacrificing quality:
| Agent | Model | Rationale |
|---|
| Specification Verifier | sonnet | Reads code + generates Gherkin - moderate reasoning |
| Test Runner | haiku | Structured execution, output parsing - low reasoning |
| Failure Analyzer | sonnet | Root cause analysis - moderate reasoning |
| Bug Fixer | opus | First-principles code fixes - high reasoning |
| Quality Gate Enforcer | haiku | Threshold comparison - low reasoning |
| Accessibility Auditor | sonnet | Code analysis + WCAG rules - moderate reasoning |
| Auto-Committer | haiku | Git operations, message formatting - low reasoning |
| Learning Optimizer | sonnet | Pattern analysis, prediction - moderate reasoning |
Projects can override model assignments in forge.config.yaml:
model_routing:
spec-verifier: sonnet
test-runner: haiku
failure-analyzer: sonnet
bug-fixer: opus
gate-enforcer: haiku
accessibility-auditor: sonnet
auto-committer: haiku
learning-optimizer: sonnet
When no override is specified, the defaults above are used. This routing reduces token cost by ~60% compared to running all agents on the highest-tier model.
Energy-Based Lane Routing
The static agent-to-model mapping above serves as the default. At runtime, Coherence Energy (E) dynamically refines routing by selecting the appropriate processing lane for each task:
| Lane | Energy Range | Processing | Latency | Description |
|---|
| Reflex | E < 0.1 | WASM engine / ruleset | < 1ms | Zero LLM calls - deterministic checks (threshold comparisons, hash validations, format checks) |
| Retrieval | 0.1 ≤ E < 0.4 | Haiku-tier + RAG | ~10ms | Pattern-matched responses with retrieval-augmented context from forge-patterns |
| Heavy | 0.4 ≤ E < 0.7 | Opus-tier deep analysis | ~100ms | First-principles reasoning for novel failures and complex fixes |
| Escalation | E ≥ 0.7 | Pause swarm | Human review | Dirichlet energy exceeds stability threshold - swarm pauses and escalates to human |
How it works: The existing UpgradeModel/DowngradeModel recommendations in criticality scoring already approximate energy-based routing. This formalizes those heuristics: when criticality is low (E < 0.1), skip the LLM entirely; when criticality exceeds the Dirichlet stability threshold (E ≥ 0.7), stop autonomous operation.
Lane selection rule: For each agent task, compute Coherence Energy E from the criticality score. The lane determines the model tier regardless of the agent's static default - a Gate Enforcer task that normally runs on haiku will escalate to opus if E ∈ [0.4, 0.7), or pause the swarm entirely if E ≥ 0.7.
PHASE 4: SPAWN AUTONOMOUS AGENTS
Claude Code MUST spawn these 8 agents in a SINGLE message with run_in_background: true:
Task({
model: "sonnet",
prompt: `You are the Specification Verifier agent. Your mission:
1. VERIFY backend is running: curl -sf http://localhost:${BACKEND_PORT}/${HEALTH_ENDPOINT}
2. Check if Gherkin specs exist for the target bounded context:
- Look in the project's spec directory
3. If specs are MISSING:
- Read the screen/component/route implementation files for the context
- Extract all user-visible features, interactions, states
- Generate Gherkin feature files with scenarios for every cyclomatic path
- Write specs to the correct location
4. If specs EXIST:
- Read current implementations
- Compare against existing scenarios
- Flag scenarios that no longer match implementation (stale specs)
- Generate new scenarios for uncovered features
- Run drift analysis: static drift (code paths vs spec steps),
contract drift (API schema vs spec expectations),
behavioral regression (historical pass/fail trends)
5. Create spec-to-test mapping:
- Each Scenario name → test function name
- Store mapping in memory for Test Runner
6. Store results:
npx @claude-flow/cli@latest memory store --key "specs-[context]-[timestamp]" \
--value "[spec status JSON]" --namespace forge-specs
CONSTRAINTS:
- NEVER generate specs for code you haven't read
- NEVER assume UI elements exist without checking implementation
- NEVER create scenarios that duplicate existing coverage
- NEVER modify existing test files - only spec files
ACCEPTANCE:
- Every implementation file has at least one Gherkin scenario
- Spec-to-test mapping has zero unmapped entries
- All generated scenarios follow Given/When/Then format
- Results stored in forge-specs namespace
Output: List of all Gherkin scenarios with their mapped test functions, and any gaps found.`,
subagent_type: "researcher",
description: "Spec Verification",
run_in_background: true
})
Task({
model: "haiku",
prompt: `You are the Test Runner agent. Your mission:
1. VERIFY backend is running
2. Check defect predictions from memory:
npx @claude-flow/cli@latest memory search --query "defect prediction [context]" --namespace forge-predictions
- Run predicted-to-fail tests FIRST for faster convergence
3. Run the E2E test suite for the specified context using the project's test command
4. Capture ALL test output including stack traces
5. Parse failures into structured format:
{testId, gherkinScenario, error, stackTrace, file, line, context}
6. Map each failure to its Gherkin scenario (from spec-to-test mapping)
7. Store results in memory for other agents:
npx @claude-flow/cli@latest memory store \
--key "test-run-[timestamp]" \
--value "[parsed results JSON]" \
--namespace forge-results
CONSTRAINTS:
- NEVER skip failing tests
- NEVER modify test code or source code
- NEVER mock internal code or stub our own APIs (external services OK)
- NEVER continue if backend health check fails
ACCEPTANCE:
- All test results stored in memory with structured format
- Zero unparsed failures - every failure has testId, error, stackTrace, file, line
- Predicted-to-fail tests executed first
- Results include Gherkin scenario mapping for every test`,
: ,
: ,
:
})
({
: ,
: ,
: ,
: ,
:
})
({
: ,
: ,
: ,
: ,
:
})
({
: ,
: ,
: ,
: ,
:
})
({
: ,
: ,
: ,
: ,
:
})
({
: ,
: ,
: ,
: ,
:
})
({
: ,
: ,
: ,
: ,
:
})
META-EVALUATION: LLM-AS-JUDGE REVIEW
After the fix loop completes, a meta-evaluation step reviews the Bug Fixer's output using a different model perspective. This catches issues that the builder misses - production evidence shows LLM-as-Judge found 0.4% test coverage across 250+ files in 7 minutes (Issue #20), while Self-Reflection caught a non-existent widget crashing 3 features (Issue #21). Multiple approaches find different issues with multiplicative (not additive) value (Issue #22).
Activation
- Automatic: When fix confidence is below Silver (< 0.75)
- Manual:
--meta-review flag on any invocation mode
- Always-on: Configurable in
forge.config.yaml with meta_review: always
Rubric (5 Dimensions)
The judge model evaluates the Bug Fixer's output against:
| Dimension | PASS Criteria | FAIL Criteria |
|---|
| Functional Completeness | No TODOs, stubs, or placeholder implementations | Any TODO/FIXME/stub found |
| Error Handling | All async operations wrapped in try-catch, user-facing error messages | Missing error handling on any API call |
| Contract Alignment | Fix matches Gherkin spec exactly, field names consistent | Spec divergence or field name mismatch |
| Existence Verification | Every widget/class/import exists in SDK/framework | Any reference to non-existent API |
| Test Quality | Tests cover happy path, error path, and edge cases | Missing error or edge case coverage |
Verdict
- PASS: All 5 dimensions satisfied - proceed to commit
- FAIL: Any dimension fails - return to Bug Fixer with specific feedback
Output
{
"verdict": "FAIL",
"dimensions": {
"functional_completeness": { "status": "PASS", "evidence": "No TODOs found" },
"error_handling": { "status": "FAIL", "evidence": "Missing try-catch on /api/v1/payments POST" },
"contract_alignment": { "status": "PASS", "evidence": "All fields match spec" },
"existence_verification": { "status": "PASS", "evidence": "All imports verified" },
"test_quality"
Anti-Echo-Chamber Guarantee
The LLM-as-Judge architecture provides a provable anti-echo-chamber property:
- Provably different priors: Bug Fixer (opus) and LLM-as-Judge (sonnet) use different model architectures with different training distributions. This guarantees their error modes are not identical.
- Error independence: If each observer has p(error) < 0.5, the probability that ALL observers make the same error is p^N (exponentially decreasing). With N=2 (Driver + Observer), P(both wrong) < 0.25.
- High-stakes escalation: For fixes below Silver confidence (< 0.75), require 3 independent priors (Driver + Observer + Failure Analyzer re-analysis). This reduces P(all wrong) < 0.125.
- Architectural diversity ceiling: The multi-tier model routing (opus/sonnet/haiku) ensures that at least 2 distinct model architectures evaluate every fix before commit.
PHASE 5: QUALITY GATES
7 gates evaluated after each fix cycle. ALL must pass before a commit is created.
| Gate | Check | Threshold | Blocking |
|---|
| 1. Functional | All tests pass | 100% pass rate | YES |
| 2. Behavioral | Gherkin scenarios satisfied | 100% of targeted scenarios | YES |
| 3. Coverage | Path coverage | >=85% overall, >=95% critical | YES (critical only) |
| 4. Security | No hardcoded secrets, secure storage, SAST checks | 0 critical/high violations | YES |
| 5. Accessibility | Accessible labels, target sizes, contrast | WCAG AA | Warning only |
| 6. Resilience | Offline handling, timeout handling, error states | Tested for target context | Warning only |
| 7. Contract | API response matches expected schema | 0 mismatches | YES |
Prime Radiant - Continuous Verification Daemon
The 7 quality gates collectively form the Prime Radiant: a continuous verification daemon that evaluates code correctness incrementally within each iteration, not just as an end-of-cycle batch.
- Streaming evaluation: Gate results update as new evidence arrives. When Test Runner completes a test batch, Gate 1 (Functional) updates immediately - it does not wait for all tests to finish. This enables early termination on blocking failures.
- Čech nerve: The bounded context open covers {U_i} form a Čech nerve N(U). Each simplex in the nerve corresponds to a set of contexts with non-empty intersection (shared dependencies). Gate 7 (Contract) evaluates on every simplex - verifying that shared types are consistent across all context overlaps.
- Inter-iteration cohomology: Between iterations, the Learning Optimizer computes global cohomology H^*(N(U)) by analyzing gate results across all contexts. A non-zero H¹ triggers immediate commit rejection and cross-context contract re-validation.
Gate Failure Categories
When gates fail, failures are categorized for targeted re-runs:
- Functional failures → Re-run Bug Fixer on failing tests
- Behavioral failures → Check spec-to-test mapping, may need new tests
- Coverage failures → Generate additional test paths
- Security failures → Fix hardcoded values, update storage patterns
- Accessibility failures → Add accessible labels, fix target sizes
- Resilience failures → Add offline/error state handling
- Contract failures → Update DTOs or flag API regression
AUTONOMOUS EXECUTION LOOP
┌────────────────────────────────────────────────────────────────────────┐
│ FORGE AUTONOMOUS LOOP │
├────────────────────────────────────────────────────────────────────────┤
│ │
│ ┌──────────┐ ┌──────────┐ ┌──────────┐ ┌──────────┐ │
│ │ Plan │───▶│ Specify │───▶│ Test │───▶│ Analyze │ │
│ │ (Approve)│ │ (Gherkin)│ │ (Run) │ │ (Root │ │
│ └──────────┘ └──────────┘ └──────────┘ │ Cause) │ │
│ ▲ └──────────┘ │
│ │ │ │
│ ┌──────────┐ ┌──────────┐ ┌──────────┐ ┌──────────┐ │
│ │ Learn │◀───│ Commit │◀───│ Gate │◀───│ Audit │ │
│ │ (Update │ │ (Auto) │ │ (7 Gates)│ │ (A11y + │ │
│ │ Tiers) │ └──────────┘ └──────────┘ │ Fix) │ │
│ └──────────┘ └──────────┘ │
│ │ │
│ └───────────────── REPEAT ──────────────────────────────────────│
│ │
│ Plan → Specify → Test → Analyze → Audit + Fix → Gate → Commit → Learn│
│ Loop continues until: ALL 7 GATES PASS or MAX_ITERATIONS (10) │
│ Gate failures are categorized for targeted re-runs (not full re-run) │
└────────────────────────────────────────────────────────────────────────┘
SPARC Pipeline Mapping: Forge's 8-phase pipeline is a refinement of the SPARC (Specification-Pseudocode-Architecture-Refinement-Completion) methodology:
| SPARC Phase | Forge Phases | Agents |
|---|
| Specification | Plan + Specify | Spec Verifier |
| Pseudocode | Test (executable specs = pseudocode made concrete) | Test Runner |
| Architecture | Analyze (root cause reveals architectural assumptions) | Failure Analyzer |
| Refinement | Fix + Gate (iterative refinement until all gates pass) | Bug Fixer, Gate Enforcer, A11y Auditor |
| Completion | Commit + Learn (verified completion with knowledge capture) | Auto-Committer, Learning Optimizer |
Forge's pipeline is a REFINEMENT of SPARC - it decomposes each SPARC phase into operationally distinct agents with quality gates between phases, enabling autonomous iteration without human intervention.
REAL-TIME PROGRESS REPORTING
Each agent emits structured progress events during execution for observability:
{"agent": "spec-verifier", "event": "spec_generated", "context": "payments", "scenarios": 12}
{"agent": "test-runner", "event": "test_started", "context": "payments", "test": "user_can_pay"}
{"agent": "test-runner", "event": "test_completed", "context": "payments", "passed": 10, "failed": 2}
{"agent": "failure-analyzer"
Progress File:
- Events are appended to
.forge/progress.jsonl (one JSON object per line)
- File is created at the start of each Forge run and truncated
- Tools can tail this file for real-time monitoring:
tail -f .forge/progress.jsonl
Integration with Agentic QE AG-UI:
- When the AQE AG-UI protocol is available, events stream directly to the user interface
- Users see live progress: which gate is being evaluated, which test is running, which fix is being applied
- When running in Claude Code without AG-UI, progress is visible through agent output files
CONFIDENCE TIERS FOR FIX PATTERNS
Every fix pattern is tracked with a confidence score that evolves over time:
{
"key": "fix-element-not-found-abc123",
"pattern": {
"error": "Element not found / No element",
"fix": "Ensure element is rendered and visible before interaction",
"files_affected": ["*_test.*"],
"context": "any"
},
"tier": "gold",
"confidence": 0.92,
"auto_apply": true,
"applied_count": 47,
"success_count": 43,
"success_rate": 0.915,
"last_applied": "2026-02-06T14:30:00Z",
"last_failed"
Tier Thresholds
| Tier | Confidence | Auto-Apply | Behavior |
|---|
| Platinum | >= 0.95 | Yes | Apply immediately without review |
| Gold | >= 0.85 | Yes | Apply and flag in commit message |
| Silver | >= 0.75 | No | Suggest to Bug Fixer, don't auto-apply |
| Bronze | >= 0.70 | No | Store for learning only, never auto-apply |
| Expired | < 0.70 | No | Pattern demoted, needs revalidation |
Confidence Updates
After each application:
- Success: confidence += 0.05 (capped at 1.0)
- Failure: confidence -= 0.10 (floored at 0.0)
- Tier promotion when crossing threshold upward
- Tier demotion when crossing threshold downward
Nash Equilibrium Property
The asymmetric update rule (+0.05 success / -0.10 failure) implements a Nash equilibrium in the pattern confidence game:
- Break-even probability: A pattern must succeed at least P(success) ≥ 2/3 ≈ 0.667 to maintain its confidence level. At exactly 2/3 success rate, expected confidence change per application = (2/3)(+0.05) + (1/3)(-0.10) = 0.
- Bronze threshold (0.70) is set just above the equilibrium point (0.667) - a pattern that barely breaks even cannot auto-apply. Only patterns with demonstrated reliability above equilibrium advance.
- No incentive exploitation: An agent cannot game the tier system by applying a pattern speculatively. The 2:1 penalty-to-reward ratio ensures that any strategy with < 67% success rate leads to demotion, making speculative application a losing strategy.
- Convergence guarantee: Patterns converge to their true success rate over time. High-quality patterns rise to Platinum; unreliable patterns sink to Expired. The equilibrium prevents oscillation.
DEFECT PREDICTION
Before running tests, the Learning Optimizer analyzes historical data to predict which tests are most likely to fail:
Input Signals
- Files changed since last green run (git diff against last-green-commit)
- Historical failure rates per bounded context (from forge-results namespace)
- Fix pattern freshness - recently applied fixes are more likely to regress
- Complexity metrics - contexts with more cyclomatic paths fail more often
- Dependency chain length - deeper dependency chains have higher failure rates
Prediction Output
{
"date": "2026-02-07",
"predictions": [
{ "context": "payments", "probability": 0.73, "reason": "3 files changed in payment module" },
{ "context": "orders", "probability": 0.45, "reason": "depends on payments (changed)" },
{ "context": "identity", "probability": 0.12, "reason": "no changes, stable history" }
],
"recommended_order": ["payments", "orders",
Tests are executed in descending probability order - predicted-to-fail tests run FIRST for faster convergence.
AGENT CRITICALITY & BOTTLENECK DETECTION
Forge continuously monitors agent performance to identify bottlenecks and optimize the orchestra. Each agent and quality gate receives a criticality score (0.0–1.0) that drives automatic optimization decisions.
Criticality Score Formula
criticality = (duration_weight × normalized_duration)
+ (blocking_weight × blocking_impact)
+ (cost_weight × normalized_cost)
+ (detection_weight × issue_detection_rate)
| Factor | Weight | Description |
|---|
| Duration | 0.30 | Wall-clock time as fraction of total run |
| Blocking Impact | 0.30 | Number of downstream agents/gates blocked while waiting |
| Model Cost | 0.20 | Token cost as fraction of total run cost |
| Issue Detection Rate | 0.20 | Ratio of real issues found to total items checked |
Bottleneck Thresholds
| Criticality | Classification | Action |
|---|
| > 0.8 | Critical bottleneck | Immediate optimization required |
| 0.5–0.8 | Moderate bottleneck | Optimization recommended |
| < 0.5 | Healthy | No action needed |
Automatic Optimization Recommendations
When a bottleneck is detected, Forge recommends (and can auto-apply) these optimizations:
| Recommendation | When Applied | Example |
|---|
| AddParallelism | Duration high, work is splittable | Run Test Runner per context in parallel |
| UpgradeModel | Detection rate low, cost is acceptable | Promote Failure Analyzer from sonnet → opus |
| DowngradeModel | Cost high, detection rate already high | Demote Gate Enforcer from sonnet → haiku |
| ReorderExecution | Blocking impact high | Move Gate 4 check earlier in the pipeline |
| CacheResults | Same analysis repeated across runs | Cache contract snapshots between runs |
Metrics Storage
{
"run_id": "forge-2026-02-19-001",
"agent_metrics": {
"bug-fixer": { "duration_ms": 45000, "cost_tokens": 12000, "issues_found": 3, "criticality": 0.72 },
"test-runner": { "duration_ms": 30000, "cost_tokens": 2000, "issues_found": 5, "criticality": 0.45 },
"gate-enforcer": { "duration_ms": 5000, "cost_tokens": 800,
Metrics are stored in the forge-criticality namespace for trend analysis across runs.
EXHAUSTIVE EDGE CASE TESTING
General UI Element Edge Cases
For EVERY interactive element, test:
-
Interaction States
- Single interaction → expected action
- Repeated rapid interaction → no duplicate action
- Long press / right-click → context menu if applicable
- Disabled state → no action, visual feedback
-
Input Field States
- Empty → placeholder visible
- Focus → visual focus indicator
- Valid input → no error
- Invalid input → error message
- Max length reached → prevents further input
- Paste → validates pasted content
- Clear → resets to empty
-
Async Operation States
- Before load → loading indicator
- During load → spinner, disabled submit
- Success → data displayed, spinner gone
- Error → error message, retry option
- Timeout → timeout message, retry option
-
Navigation Edge Cases
- Back navigation → previous screen or exit confirmation
- Deep link → correct screen with params
- Invalid deep link → fallback/error screen
- Browser forward/back (web) → correct state
-
Scroll Edge Cases
- Overscroll → appropriate feedback
- Scroll to hidden content → content becomes visible
- Keyboard appears → scroll to focused field
Network Edge Cases
- No internet → offline indicator, cached data if available
- Slow connection → loading states persist, timeout handling
- Connection restored → auto-retry pending operations
- Server error 500 → generic error message
- Auth error 401 → redirect to login
- Permission error 403 → permission denied message
- Not found 404 → "not found" message
Chaos Testing (Resilience)
For each target context, inject controlled failures:
- Timeout injection → API calls take >10s → verify timeout UI
- Partial response → API returns incomplete data → verify graceful degradation
- Rate limiting → API returns 429 → verify retry-after behavior
- Concurrent mutations → Multiple clients modify same resource → verify conflict handling
- Session expiry → Token expires mid-flow → verify re-auth prompt
Visual Regression Testing
For UI-heavy projects, Forge captures and compares screenshots to detect unintended visual changes:
- Before fix - Capture baseline screenshots of all screens in the target context
- After fix - Capture new screenshots of the same screens
- Compare - Pixel-by-pixel comparison with configurable threshold (default: 0.1% diff tolerance)
- Report - Flag visual regressions as Gate 5 (Accessibility) warnings
- Store - Save screenshot diffs in memory for review
Screenshot Capture by Platform:
| Platform | Method |
|---|
| Web (Playwright) | page.screenshot({ fullPage: true }) |
| Web (Cypress) | cy.screenshot() |
| Flutter | await tester.binding.setSurfaceSize(size); await expectLater(find.byType(App), matchesGoldenFile('name.png')) |
| Mobile (native) | Platform-specific screenshot capture |
Configuration:
visual_regression:
enabled: true
threshold: 0.001
screenshot_dir: .forge/screenshots
full_page: true
When Agentic QE is available, delegate to the visual-tester agent for parallel viewport comparison across multiple screen sizes.
Property-Based Testing
Instead of writing individual test cases, define invariants that must hold for ALL inputs. Forge extracts invariants from Gherkin specs and ADRs, then generates 1000+ random test cases per invariant.
Process:
- Extract invariants from Gherkin scenarios and ADRs (e.g., "balance is always >= 0")
- Generate random inputs covering the input space (edge values, boundary conditions, random data)
- Run each invariant against all generated inputs
- On failure: automatically shrink to the minimal counterexample
- Report the minimal failing case with full reproduction steps
Framework Tools by Language:
