@roopesh.yadava/qa-pack 1.2.0 → 1.3.0

package/README.md

@@ -31,6 +31,12 @@ After installing:
    (plus `JIRA_BASE_URL` / `JIRA_EMAIL` / `JIRA_API_TOKEN` if you want bug screenshots attached).
 2. Open `CLAUDE.md` and fill in the non-secret project facts (Jira key, environment, auth method).
 
+`.claude/settings.local.json` (seeded once from `settings.local.json.example`) auto-approves
+Playwright MCP tool calls, since those run constantly during test execution and are scoped to
+the app under test. Every Atlassian/Jira MCP call — reading a card, filing a bug, commenting,
+transitioning — and any action needing your input (filing a bug, publishing a charter,
+overwriting product context) still prompts for confirmation.
+
 Then open the repo in Claude Code and run:
 
 ```
@@ -43,8 +49,13 @@ Every run follows the same structured flow:
 
 1. **Card input** — give a Jira card ID (`run qa PROJ-123`)
 2. **Product context** — saved context + the Jira card are loaded automatically
-3. **Gap questions** — you're asked only what isn't already known (one message)
-4. **Phase selection**:
+3. **Model recommendation** — a one-line tier suggestion based on the card's Jira issue type
+   (Story/Feature/Epic → default high-capability model, Task/Bug/Sub-task → a lighter model
+   like Haiku). Informational only — switch with `/model` if you want, or ignore it; set
+   `QA_MODEL_ROUTING=off` in `.env` to skip it. See `SKILLS_CONTEXT.md` for the opt-in
+   subagent-override pattern.
+4. **Gap questions** — you're asked only what isn't already known (one message)
+5. **Phase selection**:
    - **Phase 1 — Manual testing** (UI/Figma diff → browser tests → bugs → charter)
    - **Phase 2 — Write automation** (reuse audit → Gherkin → steps → POM → real run + self-heal)
    - **Phase 3 — Both** (manual first — its findings feed the automation)

package/claude/skills/SKILLS_CONTEXT.md

@@ -210,6 +210,16 @@ Also check `token_usage_log.json` for `cache_read_input_tokens` per phase to see
 Target: `cache_efficiency_pct` ≥ 70 after the first run on a given product.
 First run will always be 0% (cold cache). Second and subsequent runs should cache the stable block (product context + skill instructions).
 
+## Model Routing (optional)
+
+`QA_MODEL_ROUTING` in `.env` controls how qa-agent responds to Jira issue type at Step 1c:
+
+| Value | Behavior |
+|-------|----------|
+| unset or `recommend` (default) | qa-agent prints a one-line model-tier recommendation (Story/Feature/Epic → high-capability, Task/Bug/Sub-task → lighter model) after fetching the card, then continues on whatever model is already active. Portable — depends on no special tooling. |
+| `off` | Skip the recommendation line entirely. |
+| `subagent` (advanced, opt-in — not wired up by the shipped skills) | For Claude Code setups where phase dispatch runs through a subagent/Task-style tool that accepts a `model` parameter: adapt qa-agent's Step 4 dispatch to pass `model: haiku` for Task/Bug and the default model for Story/Feature when invoking `manual-testing` / `automation`. This is a pattern to wire into your own environment's actual subagent tool — qa-pack itself doesn't assume one exists, since not every Claude Code setup qa-pack installs into exposes the same dispatch mechanism. |
+
 ## Artifact Locations
 
 | Artifact | Path |

package/claude/skills/manual-testing/SKILL.md

@@ -27,6 +27,14 @@ user-invocable: true
 **After the ONE allowed login snapshot — never call `browser_snapshot()` again in this run.**
 Use `browser_evaluate` with a specific CSS selector or JS expression for all DOM inspection.
 
+## Test Data & Entity Selection Rules
+
+| Situation | Forbidden | Do instead |
+|-----------|-----------|------------|
+| A test step needs a value you don't have — a user record, a transaction PIN, a one-time user code | Calling a user-list endpoint or any other app API to fetch/guess it | Ask the user once, in the normal gap-question flow ("This test needs a valid user code — please provide one, or a test user's ID"), or pull it from `.env` / product context if already recorded there |
+| A test requires two specifically related entities (e.g. a parent user + a specific child user, an account + a specific linked card) and the AC/COS doesn't name the exact pair | Picking one at random and proceeding as if it passed | Ask the user to specify the pair once. If they can't provide it inline, mark that test `🔒 BLOCKED` with the note "entity relationship not specified — needs manual verification" and continue with the rest of the plan — don't stop the whole run over one test |
+| The AC/COS doesn't say how to reach the feature under test (menu path, URL, entry point) | Reading or grepping the `src/` tree to reverse-engineer the route | Ask once: "AC doesn't say how to reach {feature} — what's the navigation path (menu/page) to get there?" |
+
 ## Token Tracking
 
 Optional and silent — follow the **Token Tracking** pattern in `SKILLS_CONTEXT.md`
@@ -152,13 +160,19 @@ HINTS = {
 
 ### 3a — Generate Test Plan
 
+Before drafting test ideas, check each AC item for a stated entry point (menu path, URL,
+button/link name). Per the **Test Data & Entity Selection Rules** above, do not scan `src/`
+to infer a missing one — collect a single navigation question per missing entry point and
+fold it into the confirmation prompt below instead of guessing.
+
 From the AC, generate numbered test ideas:
 - 1–3 tests per AC item
 - 2+ negative/edge case tests
 - 1+ error state test
 
 Show as a compact table (T-01, T-02 ... with name and expected outcome).
-Ask: `"Ready to run these tests? (yes / no or edit)"`
+Ask: `"Ready to run these tests? (yes / no or edit)"` — include any missing navigation
+questions in this same message.
 Proceed on confirmation.
 
 ### 3b — Setup Browser + Capture Login Selectors

package/claude/skills/qa-agent/SKILL.md

@@ -95,6 +95,23 @@ Criteria / COS**, **Figma link**, **issue type**, **project name** (for Step 6 f
 If the description exceeds ~800 words, keep only Title + AC/Given-When-Then blocks +
 Figma links; discard prose, comment threads, embedded images. Do not mention the truncation.
 
+### 1c — Model Recommendation
+
+Using the issue type fetched in 1b, print one line — informational only, never blocking:
+
+| Issue type | Recommendation |
+|---|---|
+| Story / Feature / Epic | high-capability model (current default) |
+| Task / Bug / Sub-task | lighter model (e.g. Haiku) |
+
+> "Issue type: {type} → {tier} recommended for this run. Switch now with `/model` if you'd
+> like, or continue on the current model."
+
+Skip this line entirely if `.env` sets `QA_MODEL_ROUTING=off`. This is a recommendation
+only — qa-pack has no way to force a mid-run model switch, so proceed on whatever model is
+active regardless of the user's choice. See `SKILLS_CONTEXT.md` → "Model Routing" for the
+opt-in subagent-override pattern some environments can wire up instead.
+
 ---
 
 ## Step 2 — Ask Relevant Questions (gaps only)

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@roopesh.yadava/qa-pack",
-  "version": "1.2.0",
+  "version": "1.3.0",
   "description": "AI-powered QA agent skills for Claude Code — manual testing, BDD automation, accessibility, UI/Figma diff, bug reporting",
   "scripts": {
     "postinstall": "node bin/postinstall.js"

package/templates/.env.example

@@ -20,3 +20,7 @@ JIRA_API_TOKEN=
 # When unset or the scripts are missing, skills skip all tracking silently.
 # QA_TRACKING_DIR=
 # QA_TRACKING_MODEL=
+
+# ── Optional: model-tier recommendation for qa-agent (see SKILLS_CONTEXT.md) ──
+# recommend (default) | off | subagent (advanced — see "Model Routing" in SKILLS_CONTEXT.md)
+# QA_MODEL_ROUTING=recommend

package/templates/settings.local.json.example

@@ -5,11 +5,7 @@
       "Bash(npx cucumber-js *)",
       "Bash(echo \"Exit: $?\")",
       "Bash(playwright --version)",
-      "mcp__claude_ai_Atlassian__getAccessibleAtlassianResources",
-      "mcp__claude_ai_Atlassian__getJiraIssue",
-      "mcp__claude_ai_Atlassian__createJiraIssue",
-      "mcp__claude_ai_Atlassian__createIssueLink",
-      "mcp__claude_ai_Atlassian__addCommentToJiraIssue"
+      "mcp__playwright"
     ]
   },
   "enableAllProjectMcpServers": true,