nw-ab-critique-dimensi…
Review dimensions for validating agent quality - template compliance, safety, testing, and priority validation
BDD requirements discovery methodology - Example Mapping, Three Amigos, conversational patterns, Given-When-Then translation, and collaborative specification
$ npx -y skills add nWave-ai/nWave --skill nw-bdd-requirements --agent claude-codeHow it fires
How this skill gets triggered: by you, by Claude, or both.
/nw-bdd-requirementsContext preview
The summary Claude sees to decide when to auto-load this skill.
BDD requirements discovery methodology - Example Mapping, Three Amigos, conversational patterns, Given-When-Then translation, and collaborative specification
name: nw-bdd-requirements description: BDD requirements discovery methodology - Example Mapping, Three Amigos, conversational patterns, Given-When-Then translation, and collaborative specification user-invocable: false disable-model-invocation: true
"If you're not having conversations, you're not doing BDD." -- Liz Keogh
BDD discovers "what we don't know we don't know" through collaborative exploration with concrete examples, not tools or formats.
Three perspectives revealing each other's blindspots:
1. **Problem Owner** (PO, BA, Domain Expert): business need | acceptance criteria | domain knowledge 2. **Problem Solver** (Developer, Engineer): technical constraints | implementation complexity | technical edge cases 3. **Skeptic** (Tester, QA): failure modes | edge cases and boundaries | challenge assumptions
1. Read user story aloud (2 min) -- shared context 2. Identify acceptance criteria/rules (8 min) -- "what must be true for done?" 3. Explore examples per rule (12 min) -- concrete scenarios, edge cases 4. Capture questions (ongoing) -- unknowns on red cards 5. Review and summarize (3 min) -- shared understanding check
If unmappable in 25 min: too large (split) | too uncertain (spike) | team needs practice.
[Yellow] User Story: Transfer money between accounts
|
+-- [Blue] Rule: Amount must not exceed source balance
| +-- [Green] $500 balance, transfer $400 -> succeeds
| +-- [Green] $500 balance, transfer $500 -> succeeds (boundary)
| +-- [Green] $500 balance, transfer $501 -> fails
| +-- [Red] What happens if balance changes during transfer?
|
+-- [Blue] Rule: Both accounts belong to same customer
| +-- [Green] Transfer between checking and savings -> succeeds
| +-- [Green] Transfer to friend's account -> requires different flow
|
+-- [Blue] Rule: Transfer creates transaction records
+-- [Green] $100 transfer -> 2 transactions (debit + credit)
+-- [Red] What timezone for timestamps?Use when: story entering sprint | multiple edge cases | team uncertain about scope | cross-functional clarification needed. Skip when: trivial well-understood story | pure technical refactoring with no behavior change | strong shared understanding already.
Template: "Is there any other context which, when this event happens, will produce a different outcome?"
Purpose: discover edge cases and alternative scenarios.
BA: "When a customer submits an order, the order is confirmed." Tester: "Is there context that produces a different outcome?" Developer: "What if the item is out of stock?" BA: "Then the order goes to backorder status." Tester: "What if payment is declined?" BA: "Then the order is pending payment."
Result: three rules discovered from one statement.
Template: "Given this context, when this event happens, is there another outcome that's important?"
BA: "When admin deletes a user account, the account is deleted." Tester: "Is there another important outcome?" Developer: "Audit log entry for the deletion." BA: "Email notification to the user." Tester: "GDPR -- delete all personal data." Developer: "What about resources owned by the user?"
Result: one simple statement revealed five important outcomes.
Template: "Can you give me a concrete example?"
Abstract rules hide assumptions. Concrete examples force decisions.
BA: "User can search products by category." Developer: "Concrete example?" BA: "User selects 'Electronics' category." Tester: "What if Electronics has 10,000 products?" BA: "We paginate. Show 20 per page." Developer: "Default sort order?" BA: "I need to ask stakeholders." [Red card: "Default sort order for category browsing?"]
Scenario titles describe WHAT the user achieves, never HOW the system implements it.
| BAD (implementation) | GOOD (business outcome) | |---|---| | FileWatcher triggers TreeView refresh | Dashboard updates in real-time | | Observer writes state.json on event | Wave progress is captured when a phase completes | | Debounce prevents rapid writes | Rapid progress changes render smoothly | | Atomic write prevents partial reads | Dashboard always sees consistent data |
**Rule**: If the title contains a class name, method name, file name, or protocol detail, rewrite it as the user-observable outcome.
Example card: "Balance $500, transfer $300 -> succeeds"
Scenario: Successful transfer with sufficient balance Given my checking account balance is $500.00 And my savings account balance is $100.00 When I transfer $300.00 from checking to savings Then my checking balance is $200.00 And my savings balance is $400.00 And I receive a confirmation message
1. Typical/happy path | 2. Boundary condition | 3. Error/alternative path
Apply Five Whys to connect stories to business value:
Story: "Add CSV export for reports" Why? "Users can export data" Why? "Analyze in Excel" Why? "No pivot tables in our app" Why? "Need to slice data by dimensions" Why? "Identify trends for business decisions" Real need: better analytical capabilities, not CSV export
Focus on behaviors contributing to business results. Challenge features la
AI agents that guide you from idea to working code, with human judgment at every gate. nWave runs inside Claude Code. It breaks feature delivery into seven waves (discover, diverge, discuss, design, devops, distill, deliver).
Repo: nWave-ai/nWave
Review dimensions for validating agent quality - template compliance, safety, testing, and priority validation
Review dimensions for validating agent quality - template compliance, safety, testing, and priority validation
Review dimensions for acceptance test quality - happy path bias, GWT compliance, business language purity, coverage completeness, walking skeleton…
Detailed 5-phase workflow for creating agents - from requirements analysis through validation and iterative refinement
5-layer testing approach for agent validation including adversarial testing, security validation, and prompt injection resistance
Architectural style selection decision matrices, trade-off analysis, structural enforcement rules, and combination patterns. Load when choosing or evaluating…