qa-engineer 0.9.1 → 0.10.0

package/skills/qa-explore/references/report-pipeline.md

@@ -11,6 +11,55 @@ For run id `R`:
 - `qa-artifacts/explore-R/explore-result.json`
 - `qa-artifacts/explore-R/screenshots/…`
 
+## Write for a reader who has never seen the product
+
+The report is forwarded. A founder, a designer, and the developer who owns one
+finding will each open it without the conversation that produced it, and often
+without knowing what the feature does. Write for that reader:
+
+- **Name things, do not point at them.** "the Sign in button on /login", not "the
+  button". A reader cannot see your screen.
+- **No unexplained jargon.** IDOR, CLS, and *soft assertion* need a clause of
+  explanation or a plainer word.
+- **State the consequence.** A reader who does not know the product cannot infer
+  why a duplicate request matters; say that it can trip rate limiting and charge
+  twice.
+
+The renderer supplies the standing furniture — what an exploratory QA pass is, what
+each dimension means, what the severity labels mean, and what could not be run. You
+supply what only this run knows: the `scope` block below, and findings whose
+`actual` and `expected` read as complete sentences.
+
+## Scope: what you did, and what you left alone
+
+Fill `scope` on every run. It is what makes the report legible to someone who was
+not there, and without it the report opens on a wall of findings.
+
+| Field | What goes in it |
+| --- | --- |
+| `objective` | One or two sentences: which feature you exercised and what you were trying to establish. Written for someone who does not know the app. |
+| `covered` | The concrete things you actually touched — screens, controls, flows, request paths. Not dimension names: the renderer already explains those. |
+| `notCovered` | Every boundary of the run, **each with its reason**. A boundary without a reason reads as an omission. |
+
+`notCovered` is the field that protects the reader. A login report that does not say
+"signing in successfully was never tested, because a QA run must not enter real
+credentials" invites the conclusion that login works. State the boundary:
+
+- what needed access or credentials you did not have, and why you did not ask
+- what a local or development environment cannot tell you (security headers, bundle
+  size, real latency) and where it must be re-checked
+- what would have caused damage to test (rate limits, lockout, destructive actions)
+- what was in scope but blocked, naming the case id
+
+Blocked cases and unrun dimensions are added by the renderer from the artifact — do
+not repeat them. Write the boundaries only it cannot know.
+
+## Order of production
+
+`explore-result.json` is written **first** and validated; the Markdown and HTML are
+renderings of it. Producing the prose first and the JSON afterwards is how the two
+drift, and the reader gets the weaker one.
+
 ## Markdown structure
 
 ```markdown
@@ -44,19 +93,51 @@ For run id `R`:
 
 Keep branding **neutral** — no org-specific colors or hosting assumptions. Local files are the product; optional publish is the user's choice.
 
-## HTML
+## JSON
 
-Produce a self-contained HTML summary:
+Write `explore-result.json` matching `contracts/explore-result.schema.json`. Validate before completion:
 
-- Simple readable typography; high-contrast text; no dependency on external CSS CDNs required for core reading.
-- Prefer inlining small screenshots as data URIs when practical; otherwise relative links to `screenshots/` beside the HTML file.
-- Include the same sections as Markdown.
+```bash
+node <SKILL_DIR>/scripts/qa-tool.mjs analysis validate \
+  qa-artifacts/explore-R/explore-result.json \
+  <SKILL_DIR>/contracts/explore-result.schema.json
+```
 
-A short standard-library Python or Node snippet written into the run folder is fine for md→html; do not require DeJoule or labs hosting.
+## HTML
 
-## JSON
+Render it from the validated JSON. Do not write it by hand and do not write a
+throwaway md→html script:
+
+```bash
+node <SKILL_DIR>/scripts/qa-tool.mjs analysis report-html \
+  qa-artifacts/explore-R/explore-result.json \
+  --out qa-artifacts/explore-R/explore-report.html
+```
 
-Write `explore-result.json` matching `contracts/explore-result.schema.json`. Validate before completion.
+The renderer produces one self-contained file — no CDN, no external stylesheet, no
+web font — with, for every finding: severity, the defect, **current behaviour**,
+**expected behaviour**, reproduction, fix direction, and each evidence entry
+(screenshots as `<img>`, everything else as a captioned excerpt). Severity ordering,
+the summary counts, the test-case table, data-validation comparisons, fix order, and
+the attribution footer come with it.
+
+So the HTML is complete only insofar as the JSON is. A finding whose `expected` says
+"should work" renders a card that says "should work". Write the fields for a reader
+who has never seen the app:
+
+| Field | Not this | This |
+| --- | --- | --- |
+| `actual` | "Login fails" | "Two identical `POST /graphql` requests are sent; the second returns 401 and the form clears" |
+| `expected` | "Should not fail" | "One request per submit; the button is disabled while in flight" |
+| `repro` | "Double-click login" | "1. Open /login  2. Enter valid credentials  3. Double-click **Sign in** within 300 ms  4. Watch the Network panel" |
+| `fixDirection` | "Fix the handler" | "Disable the submit button on the first click and re-enable it in the request's finally block" |
+
+Screenshots are referenced as paths relative to the HTML file (`screenshots/finding-01.png`),
+so the run folder stays portable as a whole.
+
+For an evidence excerpt that could carry a token or credential, run it through
+`qa-tool.mjs analysis redact` before it goes into the JSON — the renderer escapes
+markup, it does not redact secrets.
 
 ## Versioning
 

package/skills/qa-explore/scripts/qa-tool.mjs

@@ -0,0 +1,126 @@
+// synced-from: shared/tooling/qa-tool.mjs — do not edit; edit the source and run: node scripts/sync-shared.mjs --write
+// The launcher every skill invokes to reach the deterministic engine.
+//
+// A skill's SKILL.md documents exactly one command shape:
+//
+//     node <SKILL_DIR>/scripts/qa-tool.mjs <tool> <subcommand> [args]
+//
+// and this file finds the engine, wherever it happens to be. That indirection
+// exists because the pack is installed three different ways and the engine lands in
+// a different place each time:
+//
+//   1. `qa install` bundles the engine into the skill, at ./lib/. Offline, fastest,
+//      and pinned to the version that was installed.
+//   2. `npx skills add <owner>/<repo>` — or any generic file copier — copies the
+//      skill directory out of git and bundles nothing. The engine is then resolved
+//      from node_modules if the project happens to depend on the pack.
+//   3. Neither: fall back to `npx qa-engineer`, which fetches the published package
+//      on first use and is served from the npm cache afterwards.
+//
+// Because this file is committed rather than generated, path 2 works at all — which
+// is what makes the pack installable by the wider Agent Skills ecosystem. The
+// command a skill runs never changes; only where the engine came from does.
+//
+// Exit codes pass through unchanged: 0 success, 1 an invalid contract, 2 unreadable
+// input or bad usage.
+//
+// No shebang: the synced copies carry a provenance marker on line one, which would
+// sit above it and stop the kernel seeing it anyway. Every documented invocation is
+// `node qa-tool.mjs …`, which needs none.
+
+import fs from 'node:fs';
+import path from 'node:path';
+import { spawnSync } from 'node:child_process';
+import { fileURLToPath } from 'node:url';
+
+const here = path.dirname(fileURLToPath(import.meta.url));
+
+const USAGE = `usage: node qa-tool.mjs <tool> <subcommand> [args]
+
+  analysis     parse artifacts, classify errors, validate contracts, diff-guard,
+               read .qa/context.md, render an HTML report, print the footer
+  diagnostics  root cause, timeline, priority, repair plans, release readiness
+  playwright   normalize a Playwright report or summarize a trace
+
+  --where      print how the engine was resolved, and stop
+
+examples:
+  node qa-tool.mjs analysis junit test-results/results.xml
+  node qa-tool.mjs analysis report-html qa-artifacts/explore-result.json --out report.html
+  node qa-tool.mjs diagnostics report --execution-result qa-artifacts/run.json
+`;
+
+/**
+ * Where the engine is, and how we found it.
+ *
+ * Ordered by cost: a bundled copy needs no resolution and no network, a
+ * node_modules copy needs no network, and npx needs both on first use. Reporting
+ * *which* one answered matters when a skill degrades — "the engine is missing" and
+ * "the engine is being fetched" are different problems.
+ */
+function resolveEngine() {
+  const bundled = path.join(here, 'lib', 'bin', 'qa-engine.mjs');
+  if (fs.existsSync(bundled)) return { kind: 'bundled', command: [process.execPath, bundled] };
+
+  for (const base of candidateModuleRoots()) {
+    const installed = path.join(base, 'qa-engineer', 'packages', 'engine', 'bin', 'qa-engine.mjs');
+    if (fs.existsSync(installed)) {
+      return { kind: 'node_modules', command: [process.execPath, installed] };
+    }
+  }
+
+  return {
+    kind: 'npx',
+    command: [npxCommand(), '--yes', 'qa-engineer', 'engine'],
+  };
+}
+
+/** node_modules directories worth checking, nearest first. */
+function candidateModuleRoots() {
+  const roots = [];
+  let dir = here;
+  for (let depth = 0; depth < 12; depth += 1) {
+    roots.push(path.join(dir, 'node_modules'));
+    const parent = path.dirname(dir);
+    if (parent === dir) break;
+    dir = parent;
+  }
+  if (process.cwd() !== here) roots.push(path.join(process.cwd(), 'node_modules'));
+  return roots;
+}
+
+// `npx` is a shell script on POSIX and a .cmd shim on Windows; spawnSync needs the
+// exact name, and `shell: true` would put user-supplied arguments through a shell.
+function npxCommand() {
+  return process.platform === 'win32' ? 'npx.cmd' : 'npx';
+}
+
+function main(argv) {
+  if (argv.length === 0 || argv[0] === '--help' || argv[0] === '-h') {
+    process.stdout.write(USAGE);
+    return argv.length === 0 ? 2 : 0;
+  }
+
+  const engine = resolveEngine();
+
+  if (argv[0] === '--where') {
+    process.stdout.write(`${JSON.stringify({ resolved: engine.kind, command: engine.command }, null, 2)}\n`);
+    return 0;
+  }
+
+  const [program, ...prefix] = engine.command;
+  const run = spawnSync(program, [...prefix, ...argv], { stdio: 'inherit' });
+
+  if (run.error) {
+    // Say which path was tried and what to do, because a skill's fallback prose
+    // cannot diagnose this and the user is the one who has to fix it.
+    const advice = engine.kind === 'npx'
+      ? 'the engine is not bundled and npx is unavailable — run `npx qa-engineer install` in this project, or install Node 18+'
+      : `could not execute ${program}`;
+    process.stderr.write(`qa-tool: ${advice}\n${run.error.message}\n`);
+    return 2;
+  }
+  return run.status ?? 2;
+}
+
+process.exitCode = main(process.argv.slice(2));

package/skills/qa-fix/SKILL.md

@@ -54,14 +54,14 @@ This milestone plans repairs; it does not write or apply code. It never edits a
 
 ## Tooling
 
-Invoke the bundled engine through its launcher, as documented in [references/deterministic-tooling.md](references/deterministic-tooling.md). `SKILL_DIR` below is this skill's own directory — `.agents/skills/qa-fix` or `.claude/skills/qa-fix`, whichever exists. The command shape is the same in bash, zsh, PowerShell, and cmd.exe; on Windows use `python` if `python3` is not on PATH.
+Invoke the bundled engine through its launcher, as documented in [references/deterministic-tooling.md](references/deterministic-tooling.md). `SKILL_DIR` below is this skill's own directory — `.agents/skills/qa-fix` or `.claude/skills/qa-fix`, whichever exists. The command shape is the same in bash, zsh, PowerShell, and cmd.exe, and it runs under the same Node that installed the pack — there is no second runtime to find.
 
 | Tool | Invocation | Output | Fallback |
 | --- | --- | --- | --- |
-| Repair planner | `python3 <SKILL_DIR>/scripts/qa_tool.py diagnostics plan-repairs --diagnosis <path>` | A repair plan per diagnosis entry, escalations included | Reason over the repair-strategy module manually and mark the plan degraded |
-| Diff guard | `python3 <SKILL_DIR>/scripts/qa_tool.py analysis diff-guard <diff-file>` | `{issues, safe}` — `safe:false` means the change is unsafe | None: without the guard, record `diffGuardReview.status` as `not-run` and never claim a diff is safe |
+| Repair planner | `node <SKILL_DIR>/scripts/qa-tool.mjs diagnostics plan-repairs --diagnosis <path>` | A repair plan per diagnosis entry, escalations included | Reason over the repair-strategy module manually and mark the plan degraded |
+| Diff guard | `node <SKILL_DIR>/scripts/qa-tool.mjs analysis diff-guard <diff-file>` | `{issues, safe}` — `safe:false` means the change is unsafe | None: without the guard, record `diffGuardReview.status` as `not-run` and never claim a diff is safe |
 
-A missing `qa_tool.py` means the engine is not installed.
+A missing `qa-tool.mjs` means the engine is not installed.
 
 **The diff guard is not advisory.** Whenever a diff exists — drafted here or supplied by the user — run it through the guard and record the verdict in `diffGuardReview`. A `fail` verdict forbids the disposition `repairable`; the contract rejects that combination, so escalate instead. A plan that carries no diff records `not-run` and claims nothing about safety.
 

package/skills/qa-fix/references/deterministic-tooling.md

@@ -17,7 +17,7 @@ resolves its own location. There is nothing to set up and no shell features are
 involved:
 
 ```bash
-python3 <skill-dir>/scripts/qa_tool.py <tool> <subcommand> [args]
+node <skill-dir>/scripts/qa-tool.mjs <tool> <subcommand> [args]
 ```
 
 `<skill-dir>` is wherever the host installed this skill — usually
@@ -25,19 +25,21 @@ python3 <skill-dir>/scripts/qa_tool.py <tool> <subcommand> [args]
 whichever exists.
 
 ```bash
-python3 .agents/skills/qa-run/scripts/qa_tool.py analysis junit test-results/results.xml
+node .agents/skills/qa-run/scripts/qa-tool.mjs analysis junit test-results/results.xml
 ```
 
-That line is identical in bash, zsh, PowerShell, and cmd.exe. **On Windows, use
-`python` if `python3` is not on PATH** — that is the only platform difference.
+That line is identical in bash, zsh, PowerShell, and cmd.exe, with no platform
+difference at all: it runs under the same Node the user already had to install the
+pack.
 
-An earlier version of this contract used a shell recipe
-(`QA_LIB="$(ls -d … | head -1)"` with a `PYTHONPATH=` prefix). It was POSIX-only,
-so on Windows every deterministic call failed, each skill fell back to its manual
-path, and the user silently got guesswork while believing the tooling had run.
-Never reintroduce a shell-dependent invocation.
+Two earlier versions of this contract failed the same way. The first was a shell
+recipe (`QA_LIB="$(ls -d … | head -1)"` with a `PYTHONPATH=` prefix), POSIX-only, so
+on Windows every deterministic call failed and each skill fell back to guesswork
+while appearing to have run its tooling. The second was portable but needed a Python
+interpreter the user never agreed to install. Never reintroduce a shell-dependent
+invocation, and never add a runtime the install did not already require.
 
-If `qa_tool.py` is missing, the engine is not installed: say so, recommend
+If `qa-tool.mjs` is missing, the engine is not installed: say so, recommend
 `qa repair`, use the skill's documented fallback, and mark the result degraded.
 
 Every tool writes JSON to stdout. Exit `0` means success; exit `1` means an
@@ -45,33 +47,35 @@ invalid contract; exit `2` means unreadable input, a malformed artifact, or a
 payload that failed its seam contract, and the JSON body carries `error` and
 `detail`. Treat a non-zero exit as missing evidence, never as a value to guess.
 
-Standard-library Python 3.8+ only — nothing to install.
+Dependency-free Node — nothing to install beyond the Node that ran `npx`.
 
-## 2. Analysis core — `qa_tool.py analysis`
+## 2. Analysis core — `qa-tool.mjs analysis`
 
 Framework-agnostic parsing, redaction, and validation.
 
 | Subcommand | Invocation | Returns |
 | --- | --- | --- |
-| `junit` | `python3 <skill-dir>/scripts/qa_tool.py analysis junit <report.xml>` | `{tests: {...}, executed: [...]}` normalized counts and per-test outcomes |
-| `har` | `python3 <skill-dir>/scripts/qa_tool.py analysis har <file.har> [--slow-ms N]` | Redacted request/response summary, failures, slow calls |
-| `discover` | `python3 <skill-dir>/scripts/qa_tool.py analysis discover [--root DIR] [--path P]` | Artifacts found, by type, with presence flags |
-| `diff-guard` | `python3 <skill-dir>/scripts/qa_tool.py analysis diff-guard <diff-file>` | `{issues: [...], safe: bool}` — `safe:false` blocks the change |
-| `redact` | `python3 <skill-dir>/scripts/qa_tool.py analysis redact <file>` | The file's text with credentials masked |
-| `validate` | `python3 <skill-dir>/scripts/qa_tool.py analysis validate <instance.json> <schema.json>` | `{valid: bool, errors: [...]}`; exit 1 when invalid |
-| `classify` | `python3 <skill-dir>/scripts/qa_tool.py analysis classify "<error message>" [--http-status N]` | `{classification, confidence, reason}` from the shared taxonomy |
-| `context` | `python3 <skill-dir>/scripts/qa_tool.py analysis context [--root DIR] [--path .qa/context.md]` | The parsed, schema-validated project context as JSON |
-
-## 3. Diagnostic engine — `qa_tool.py diagnostics`
+| `junit` | `node <skill-dir>/scripts/qa-tool.mjs analysis junit <report.xml>` | `{tests: {...}, executed: [...]}` normalized counts and per-test outcomes |
+| `har` | `node <skill-dir>/scripts/qa-tool.mjs analysis har <file.har> [--slow-ms N]` | Redacted request/response summary, failures, slow calls |
+| `discover` | `node <skill-dir>/scripts/qa-tool.mjs analysis discover [--root DIR] [--path P]` | Artifacts found, by type, with presence flags |
+| `diff-guard` | `node <skill-dir>/scripts/qa-tool.mjs analysis diff-guard <diff-file>` | `{issues: [...], safe: bool}` — `safe:false` blocks the change |
+| `redact` | `node <skill-dir>/scripts/qa-tool.mjs analysis redact <file>` | The file's text with credentials masked |
+| `validate` | `node <skill-dir>/scripts/qa-tool.mjs analysis validate <instance.json> <schema.json>` | `{valid: bool, errors: [...]}`; exit 1 when invalid |
+| `classify` | `node <skill-dir>/scripts/qa-tool.mjs analysis classify "<error message>" [--http-status N]` | `{classification, confidence, reason}` from the shared taxonomy |
+| `context` | `node <skill-dir>/scripts/qa-tool.mjs analysis context [--root DIR] [--path .qa/context.md]` | The parsed, schema-validated project context as JSON |
+| `report-html` | `node <skill-dir>/scripts/qa-tool.mjs analysis report-html <result.json> [--out report.html]` | The result rendered as one self-contained HTML report — every required field, footer included. Run it instead of writing HTML |
+| `branding` | `node <skill-dir>/scripts/qa-tool.mjs analysis branding --format markdown\|html\|text` | The exact attribution footer bytes for a **rendered** report (never for a JSON artifact) |
+
+## 3. Diagnostic engine — `qa-tool.mjs diagnostics`
 
 One engine, consumed by the diagnostic skills. Reasoning lives here once.
 
 | Subcommand | Invocation | Returns |
 | --- | --- | --- |
-| `diagnose` | `python3 <skill-dir>/scripts/qa_tool.py diagnostics diagnose --execution-result <path> [--analysis-result <path>]` | `{entries: [...], timeline: [...], recommendations: [...]}` |
-| `plan-repairs` | `python3 <skill-dir>/scripts/qa_tool.py diagnostics plan-repairs --diagnosis <path>` | `{plans: [...]}` — one plan per entry, escalations included |
-| `summarize` | `python3 <skill-dir>/scripts/qa_tool.py diagnostics summarize --execution-result <path> --diagnosis <path>` | `{totals, byClassification, topPriority, releaseReadiness}` |
-| `report` | `python3 <skill-dir>/scripts/qa_tool.py diagnostics report --execution-result <path> [--analysis-result <path>]` | `{diagnosis, plans, summary}` — all three in one call |
+| `diagnose` | `node <skill-dir>/scripts/qa-tool.mjs diagnostics diagnose --execution-result <path> [--analysis-result <path>]` | `{entries: [...], timeline: [...], recommendations: [...]}` |
+| `plan-repairs` | `node <skill-dir>/scripts/qa-tool.mjs diagnostics plan-repairs --diagnosis <path>` | `{plans: [...]}` — one plan per entry, escalations included |
+| `summarize` | `node <skill-dir>/scripts/qa-tool.mjs diagnostics summarize --execution-result <path> --diagnosis <path>` | `{totals, byClassification, topPriority, releaseReadiness}` |
+| `report` | `node <skill-dir>/scripts/qa-tool.mjs diagnostics report --execution-result <path> [--analysis-result <path>]` | `{diagnosis, plans, summary}` — all three in one call |
 
 **Inputs.** `--execution-result` takes a `qa-run` execution result, or the minimal
 subset (`tests` counts plus `executed[]` entries carrying `status`).
@@ -99,7 +103,7 @@ contract names:
 
 This mapping is not busywork: the strictness is what stops a skill from shipping a
 result whose shape nobody checked. Validate before completion —
-`python3 <skill-dir>/scripts/qa_tool.py analysis validate <result.json> <schema.json>` — and fix the
+`node <skill-dir>/scripts/qa-tool.mjs analysis validate <result.json> <schema.json>` — and fix the
 result, never the claim.
 
 ## 4. Framework adapters
@@ -109,11 +113,11 @@ a `--framework` flag.
 
 | Adapter | Invocation | Returns |
 | --- | --- | --- |
-| Playwright report | `python3 <skill-dir>/scripts/qa_tool.py playwright report <results.json>` | The same `{tests, executed}` shape as `junit` |
-| Playwright trace | `python3 <skill-dir>/scripts/qa_tool.py playwright trace <trace.zip>` | Actions, console/network counts, errors, classification |
+| Playwright report | `node <skill-dir>/scripts/qa-tool.mjs playwright report <results.json>` | The same `{tests, executed}` shape as `junit` |
+| Playwright trace | `node <skill-dir>/scripts/qa-tool.mjs playwright trace <trace.zip>` | Actions, console/network counts, errors, classification |
 
 For Selenium, Cypress, and WebdriverIO, normalize through
-`qa_tool.py analysis junit` — those adapters have no richer artifact than JUnit, and
+`qa-tool.mjs analysis junit` — those adapters have no richer artifact than JUnit, and
 the skill says so rather than implying trace-grade depth.
 
 ## 5. Reporting what ran

package/skills/qa-fix/references/diagnostic-engine.md

@@ -5,11 +5,11 @@ How the one engine composes the other platforms into a diagnosis, and what lives
 
 ## What the engine does, in code
 
-The `qa_diagnostics` package (in lib/) implements the deterministic reasoning, reusing `qa_analysis`:
+The `diagnostics` package (in lib/) implements the deterministic reasoning, reusing `analysis`:
 
 | Step | Module | Reuses |
 | --- | --- | --- |
-| Classify the failure into a root cause | `root_cause` | `qa_analysis.taxonomy` |
+| Classify the failure into a root cause | `root_cause` | `the shared taxonomy` |
 | Assign severity, priority, impacts, owner, effort | `prioritization` | the taxonomy classes |
 | Reconstruct the ordered timeline | `timeline` | the evidence in findings |
 | Plan a repair (never code) | `repair` | the taxonomy; the diff guard's guarantees |
@@ -33,4 +33,4 @@ The engine calls the analysis platform for classification and evidence, the exec
 
 ## Bundling
 
-Because the diagnostic skills run the engine in a consumer's repository, the `qa_analysis` and `qa_diagnostics` packages are bundled into each skill's `scripts/lib/` from their canonical source in `shared/`. The bundle is a build artifact produced by the bundler; the source of truth is `shared/`. See ADR-0011.
+Because the diagnostic skills run the engine in a consumer's repository, the `analysis` and `diagnostics` packages are bundled into each skill's `scripts/lib/` from their canonical source in `shared/`. The bundle is a build artifact produced by the bundler; the source of truth is `shared/`. See ADR-0011.

package/skills/qa-fix/references/evidence-and-reporting.md

@@ -27,17 +27,42 @@ Any skill that reaches a conclusion — a detected fact, a chosen strategy, a cl
 | Confidence | Calibrated per rule 4; omitted rather than invented when the skill did not weigh alternatives |
 | Recommendations | The next action, named concretely — often the next command and the artifact to feed it |
 
+## Rendering the HTML report
+
+**Never type the HTML.** When a skill's result conforms to a contract the renderer
+supports, the HTML report is generated from that artifact:
+
+```bash
+node <SKILL_DIR>/scripts/qa-tool.mjs analysis report-html <result.json> --out <report.html>
+```
+
+The reason is not tidiness. A hand-written report is a second, lossy copy of the
+artifact, and it loses exactly what the reader needs: the first live `/qa-explore`
+run wrote findings whose `actual`, `expected`, and `fixDirection` were all required,
+all present in the JSON, and all absent from the page — collapsed into one sentence
+that left the reader to guess what correct behaviour would have been. The renderer
+cannot make that mistake, because the fields are in its template.
+
+What the renderer guarantees per finding: severity, the defect, **what happens
+now**, **what should happen instead**, how to reproduce it, the fix direction, and
+every evidence entry — plus the attribution footer, in one self-contained file with
+no external assets.
+
+Run `qa-tool.mjs analysis report-html --help` for the supported contracts. For an
+artifact it does not support, write the HTML by hand from the contract's fields —
+and render **every** field a reader needs, in that order.
+
 ## Attribution on rendered reports
 
 A report a person opens carries a product attribution footer, the way a Lighthouse
 or Allure report does. It is rendered, never typed: the exact bytes come from the
 bundled analysis toolkit, so every report is identical and a wording change is a
-one-file edit.
+one-file edit. `report-html` already embeds it; for any other rendering, ask for it:
 
 ```bash
-PYTHONPATH="$QA_LIB" python3 -m qa_analysis.cli branding --format html
-PYTHONPATH="$QA_LIB" python3 -m qa_analysis.cli branding --format markdown
-PYTHONPATH="$QA_LIB" python3 -m qa_analysis.cli branding --format text
+node <SKILL_DIR>/scripts/qa-tool.mjs analysis branding --format markdown
+node <SKILL_DIR>/scripts/qa-tool.mjs analysis branding --format html
+node <SKILL_DIR>/scripts/qa-tool.mjs analysis branding --format text
 ```
 
 Append the output as the last element of the rendered document: `html` inside

package/skills/qa-fix/scripts/qa-tool.mjs

@@ -0,0 +1,126 @@
+// synced-from: shared/tooling/qa-tool.mjs — do not edit; edit the source and run: node scripts/sync-shared.mjs --write
+// The launcher every skill invokes to reach the deterministic engine.
+//
+// A skill's SKILL.md documents exactly one command shape:
+//
+//     node <SKILL_DIR>/scripts/qa-tool.mjs <tool> <subcommand> [args]
+//
+// and this file finds the engine, wherever it happens to be. That indirection
+// exists because the pack is installed three different ways and the engine lands in
+// a different place each time:
+//
+//   1. `qa install` bundles the engine into the skill, at ./lib/. Offline, fastest,
+//      and pinned to the version that was installed.
+//   2. `npx skills add <owner>/<repo>` — or any generic file copier — copies the
+//      skill directory out of git and bundles nothing. The engine is then resolved
+//      from node_modules if the project happens to depend on the pack.
+//   3. Neither: fall back to `npx qa-engineer`, which fetches the published package
+//      on first use and is served from the npm cache afterwards.
+//
+// Because this file is committed rather than generated, path 2 works at all — which
+// is what makes the pack installable by the wider Agent Skills ecosystem. The
+// command a skill runs never changes; only where the engine came from does.
+//
+// Exit codes pass through unchanged: 0 success, 1 an invalid contract, 2 unreadable
+// input or bad usage.
+//
+// No shebang: the synced copies carry a provenance marker on line one, which would
+// sit above it and stop the kernel seeing it anyway. Every documented invocation is
+// `node qa-tool.mjs …`, which needs none.
+
+import fs from 'node:fs';
+import path from 'node:path';
+import { spawnSync } from 'node:child_process';
+import { fileURLToPath } from 'node:url';
+
+const here = path.dirname(fileURLToPath(import.meta.url));
+
+const USAGE = `usage: node qa-tool.mjs <tool> <subcommand> [args]
+
+  analysis     parse artifacts, classify errors, validate contracts, diff-guard,
+               read .qa/context.md, render an HTML report, print the footer
+  diagnostics  root cause, timeline, priority, repair plans, release readiness
+  playwright   normalize a Playwright report or summarize a trace
+
+  --where      print how the engine was resolved, and stop
+
+examples:
+  node qa-tool.mjs analysis junit test-results/results.xml
+  node qa-tool.mjs analysis report-html qa-artifacts/explore-result.json --out report.html
+  node qa-tool.mjs diagnostics report --execution-result qa-artifacts/run.json
+`;
+
+/**
+ * Where the engine is, and how we found it.
+ *
+ * Ordered by cost: a bundled copy needs no resolution and no network, a
+ * node_modules copy needs no network, and npx needs both on first use. Reporting
+ * *which* one answered matters when a skill degrades — "the engine is missing" and
+ * "the engine is being fetched" are different problems.
+ */
+function resolveEngine() {
+  const bundled = path.join(here, 'lib', 'bin', 'qa-engine.mjs');
+  if (fs.existsSync(bundled)) return { kind: 'bundled', command: [process.execPath, bundled] };
+
+  for (const base of candidateModuleRoots()) {
+    const installed = path.join(base, 'qa-engineer', 'packages', 'engine', 'bin', 'qa-engine.mjs');
+    if (fs.existsSync(installed)) {
+      return { kind: 'node_modules', command: [process.execPath, installed] };
+    }
+  }
+
+  return {
+    kind: 'npx',
+    command: [npxCommand(), '--yes', 'qa-engineer', 'engine'],
+  };
+}
+
+/** node_modules directories worth checking, nearest first. */
+function candidateModuleRoots() {
+  const roots = [];
+  let dir = here;
+  for (let depth = 0; depth < 12; depth += 1) {
+    roots.push(path.join(dir, 'node_modules'));
+    const parent = path.dirname(dir);
+    if (parent === dir) break;
+    dir = parent;
+  }
+  if (process.cwd() !== here) roots.push(path.join(process.cwd(), 'node_modules'));
+  return roots;
+}
+
+// `npx` is a shell script on POSIX and a .cmd shim on Windows; spawnSync needs the
+// exact name, and `shell: true` would put user-supplied arguments through a shell.
+function npxCommand() {
+  return process.platform === 'win32' ? 'npx.cmd' : 'npx';
+}
+
+function main(argv) {
+  if (argv.length === 0 || argv[0] === '--help' || argv[0] === '-h') {
+    process.stdout.write(USAGE);
+    return argv.length === 0 ? 2 : 0;
+  }
+
+  const engine = resolveEngine();
+
+  if (argv[0] === '--where') {
+    process.stdout.write(`${JSON.stringify({ resolved: engine.kind, command: engine.command }, null, 2)}\n`);
+    return 0;
+  }
+
+  const [program, ...prefix] = engine.command;
+  const run = spawnSync(program, [...prefix, ...argv], { stdio: 'inherit' });
+
+  if (run.error) {
+    // Say which path was tried and what to do, because a skill's fallback prose
+    // cannot diagnose this and the user is the one who has to fix it.
+    const advice = engine.kind === 'npx'
+      ? 'the engine is not bundled and npx is unavailable — run `npx qa-engineer install` in this project, or install Node 18+'
+      : `could not execute ${program}`;
+    process.stderr.write(`qa-tool: ${advice}\n${run.error.message}\n`);
+    return 2;
+  }
+  return run.status ?? 2;
+}
+
+process.exitCode = main(process.argv.slice(2));

package/skills/qa-flaky/SKILL.md

@@ -53,14 +53,14 @@ Do not use it to run tests (`/qa-run`) or to fix a deterministic failure (`/qa-d
 
 ## Tooling
 
-Invoke the bundled engine through its launcher, as documented in [references/deterministic-tooling.md](references/deterministic-tooling.md). `SKILL_DIR` below is this skill's own directory — `.agents/skills/qa-flaky` or `.claude/skills/qa-flaky`, whichever exists. The command shape is the same in bash, zsh, PowerShell, and cmd.exe; on Windows use `python` if `python3` is not on PATH.
+Invoke the bundled engine through its launcher, as documented in [references/deterministic-tooling.md](references/deterministic-tooling.md). `SKILL_DIR` below is this skill's own directory — `.agents/skills/qa-flaky` or `.claude/skills/qa-flaky`, whichever exists. The command shape is the same in bash, zsh, PowerShell, and cmd.exe, and it runs under the same Node that installed the pack — there is no second runtime to find.
 
 | Tool | Invocation | Output | Fallback |
 | --- | --- | --- | --- |
-| Diagnostic engine | `python3 <SKILL_DIR>/scripts/qa_tool.py diagnostics diagnose --execution-result <path>` | Deterministic flaky classification and prioritized causes (retry counts drive the `flaky` classification) | Reason over the flakiness module manually and mark the verdict degraded |
-| JUnit normalizer | `python3 <SKILL_DIR>/scripts/qa_tool.py analysis junit <report.xml>` | Per-test retries and final status across runs — the input the flake rate is computed from | Read the reporter and state that the flake rate is unmeasured |
+| Diagnostic engine | `node <SKILL_DIR>/scripts/qa-tool.mjs diagnostics diagnose --execution-result <path>` | Deterministic flaky classification and prioritized causes (retry counts drive the `flaky` classification) | Reason over the flakiness module manually and mark the verdict degraded |
+| JUnit normalizer | `node <SKILL_DIR>/scripts/qa-tool.mjs analysis junit <report.xml>` | Per-test retries and final status across runs — the input the flake rate is computed from | Read the reporter and state that the flake rate is unmeasured |
 
-A missing `qa_tool.py` means the engine is not installed. A flake rate is computed from observed runs or reported as `null` — never estimated.
+A missing `qa-tool.mjs` means the engine is not installed. A flake rate is computed from observed runs or reported as `null` — never estimated.
 
 ## Output