ccqa 1.10.0 → 1.10.1

package/dist/bin/ccqa.mjs

@@ -39,6 +39,26 @@ const EVIDENCE_SUBDIR = "evidence";
 /** Per-spec run artifacts for external (runCommand) targets: `artifacts/<feature>__<spec>/`. */
 const ARTIFACTS_SUBDIR = "artifacts";
 //#endregion
+//#region src/run/errors.ts
+/**
+* Usage error (bad flag combination, broken profile, failed `git diff`, …)
+* thrown by the `run` pipeline and the helpers it calls, e.g.
+* `collectChangedSpecs`. `executeRun` never calls `process.exit`, so each
+* host maps this itself: the CLI action catches it and exits with
+* `exitCode`; the hub runner records it as a run-level error.
+*/
+var RunUsageError = class extends Error {
+	exitCode = 2;
+	constructor(message) {
+		super(message);
+		this.name = "RunUsageError";
+	}
+};
+/** An error's message, whatever was thrown. Shared so the run modules report failures alike. */
+function errMessage(err) {
+	return err instanceof Error ? err.message : String(err);
+}
+//#endregion
 //#region src/runtime/env-vars.ts
 const ENV_VAR_RE = /\$\{([A-Z_][A-Z0-9_]*)\}|\$([A-Z_][A-Z0-9_]*)/g;
 const ANY_VAR_RE = /\$\{([A-Za-z_][A-Za-z0-9_]*)\}|\$([A-Za-z_][A-Za-z0-9_]*)/g;
@@ -399,7 +419,7 @@ function parseSpecPath(specPath) {
 		featureName: parts[0],
 		specName: parts[1]
 	};
-	throw new Error(`Invalid spec path: "${specPath}". Expected "<feature>/<spec>" or "features/<feature>/test-cases/<spec>".`);
+	throw new RunUsageError(`Invalid spec path: "${specPath}". Expected "<feature>/<spec>" or "features/<feature>/test-cases/<spec>".`);
 }
 function getFeatureDir(featureName, cwd) {
 	return join(getCcqaDir(cwd), "features", featureName);
@@ -4669,6 +4689,34 @@ function isWithin(rootAbs, abs) {
 	return abs === rootAbs || abs.startsWith(rootAbs + sep);
 }
 //#endregion
+//#region src/cli/usage-errors.ts
+/**
+* Turn a `RunUsageError` into `[error] <message>` plus its exit code, for a
+* commander action.
+*
+* The helpers shared across commands — `resolveAnalysisBase`,
+* `collectChangedSpecs` — signal a bad invocation this way, so every command
+* that calls one needs the same boundary. Without it the process dies on an
+* unhandled rejection and prints a stack trace: `bin/ccqa.ts` installs no
+* global handler, so there is nowhere else for it to land.
+*
+* Lives here rather than beside `RunUsageError` because `src/run/errors.ts` is
+* deliberately dependency-free, and this needs the logger.
+*/
+function withUsageErrors(fn) {
+	return async (...args) => {
+		try {
+			await fn(...args);
+		} catch (err) {
+			if (err instanceof RunUsageError) {
+				error(err.message);
+				process.exit(err.exitCode);
+			}
+			throw err;
+		}
+	};
+}
+//#endregion
 //#region src/runtime/profile-env.ts
 /**
 * Profile env vars are hub-sourced (pulled via `ccqa hub`) and merged into
@@ -5121,7 +5169,7 @@ const DraftNamingSchema = z.object({
 //#endregion
 //#region src/cli/draft.ts
 const CATEGORY_LABEL = DRAFT_CATEGORY_LABEL;
-const draftCommand = addLanguageOption(new Command("draft").argument("[feature/spec]", "Optional spec path (e.g. tasks/create-and-complete). If omitted, Claude proposes one from your intent.").description("Interactively draft and refine a spec.yaml with Claude Code").option("--instruction <text>", "Non-interactive single-shot instruction (skips the interactive loop)").option("--apply", "Auto-apply each generated patch without [y/N] confirmation", false)).action(async (specPath, opts) => {
+const draftCommand = addLanguageOption(new Command("draft").argument("[feature/spec]", "Optional spec path (e.g. tasks/create-and-complete). If omitted, Claude proposes one from your intent.").description("Interactively draft and refine a spec.yaml with Claude Code").option("--instruction <text>", "Non-interactive single-shot instruction (skips the interactive loop)").option("--apply", "Auto-apply each generated patch without [y/N] confirmation", false)).action(withUsageErrors(async (specPath, opts) => {
 	await ensureCcqaDir();
 	let featureName;
 	let specName;
@@ -5134,7 +5182,7 @@ const draftCommand = addLanguageOption(new Command("draft").argument("[feature/s
 		prefilledIntent = intent;
 	}
 	await runDraft(featureName, specName, opts, prefilledIntent);
-});
+}));
 async function runDraft(featureName, specName, opts, prefilledIntent) {
 	header("draft", `${featureName}/${specName}`);
 	const ja = useJapanesePrompts(opts.language);
@@ -6426,26 +6474,6 @@ function createDiffProvider(args) {
 	} };
 }
 //#endregion
-//#region src/run/errors.ts
-/**
-* Usage error (bad flag combination, broken profile, failed `git diff`, …)
-* thrown by the `run` pipeline and the helpers it calls, e.g.
-* `collectChangedSpecs`. `executeRun` never calls `process.exit`, so each
-* host maps this itself: the CLI action catches it and exits with
-* `exitCode`; the hub runner records it as a run-level error.
-*/
-var RunUsageError = class extends Error {
-	exitCode = 2;
-	constructor(message) {
-		super(message);
-		this.name = "RunUsageError";
-	}
-};
-/** An error's message, whatever was thrown. Shared so the run modules report failures alike. */
-function errMessage(err) {
-	return err instanceof Error ? err.message : String(err);
-}
-//#endregion
 //#region src/run/git-context.ts
 /** The `--failure-analysis` value that selects per-spec hub-ledger baselines. */
 const LAST_GREEN = "last-green";
@@ -8214,6 +8242,14 @@ function isCcqaPath(path) {
 	return /(?:^|\/)\.ccqa\//.test(path);
 }
 /**
+* A malformed reply costs the whole selection, so it is worth one more call
+* before giving up. `ccqa drift` retries per spec for the same reason; this
+* call carries every undecided spec at once, so the blast radius is larger,
+* not smaller. Observed in practice: three runs over one commit produced a
+* parse failure, a clean answer, and a different clean answer.
+*/
+const MAX_ATTEMPTS = 2;
+/**
 * One model call for the whole undecided set, not one per spec: the specs are
 * judged against the same diff, and seeing them together is what lets the
 * model tell them apart.
@@ -8224,32 +8260,44 @@ function isCcqaPath(path) {
 */
 async function judgeWithModel(input) {
 	const { productChanges, undecided, cwd, base, head, model } = input;
-	const { result, isError } = await invokeClaudeStreaming({
-		prompt: buildSelectPrompt({
-			changed: productChanges,
-			specs: undecided,
-			base,
-			head
-		}),
-		systemPrompt: buildSelectSystemPrompt(),
-		allowedTools: [
-			"Read",
-			"Grep",
-			"Glob"
-		],
-		silenceBashLog: true,
-		cwd,
-		...model ? { model } : {}
-	}, (_msg) => {});
-	if (isError) return abandonSelection(undecided, "the selection model returned an error");
-	const json = extractJsonBlock(result);
-	if (!json) return abandonSelection(undecided, "the selection model returned no JSON block");
 	let parsed;
-	try {
-		parsed = JSON.parse(json);
-	} catch (e) {
-		return abandonSelection(undecided, `the selection model's JSON did not parse: ${e.message}`);
+	let lastError = "";
+	for (let attempt = 1; attempt <= MAX_ATTEMPTS; attempt++) {
+		const { result, isError } = await invokeClaudeStreaming({
+			prompt: buildSelectPrompt({
+				changed: productChanges,
+				specs: undecided,
+				base,
+				head
+			}),
+			systemPrompt: buildSelectSystemPrompt(),
+			allowedTools: [
+				"Read",
+				"Grep",
+				"Glob"
+			],
+			silenceBashLog: true,
+			cwd,
+			...model ? { model } : {}
+		}, (_msg) => {});
+		if (isError) {
+			lastError = "the selection model returned an error";
+			continue;
+		}
+		const json = extractJsonBlock(result);
+		if (!json) {
+			lastError = "the selection model returned no JSON block";
+			continue;
+		}
+		try {
+			parsed = JSON.parse(json);
+			lastError = "";
+			break;
+		} catch (e) {
+			lastError = `the selection model's JSON did not parse: ${e.message}`;
+		}
 	}
+	if (lastError) return abandonSelection(undecided, `${lastError} (${MAX_ATTEMPTS} attempts)`);
 	const changedPaths = new Set(productChanges.map((f) => f.path));
 	const byUndecidedKey = new Map(undecided.map((s) => [specKey(s), s]));
 	const answers = /* @__PURE__ */ new Map();
@@ -14954,7 +15002,7 @@ async function confirmOverwrite(path) {
 		rl.close();
 	}
 }
-const generateCommand = addHubOptions(addProfileOption(addLanguageOption(new Command("generate").argument("<feature/spec>", "Spec id in '<feature>/<spec>' form (resolves to .ccqa/features/<feature>/test-cases/<spec>/)").description("Generate test code from a spec via its target plugin. Recording-backed targets compile the existing ir.json (run `ccqa record` first); spec-input targets generate directly from the spec.").option("-m, --model <name>", "Claude model alias ('sonnet'|'opus'|'haiku') or full ID. Overrides CCQA_MODEL.").option("--target <id>", "Generate through this target instead of the spec's own — e.g. emit a Playwright spec from an agent-browser recording. The spec's `target:` stays the default for `ccqa run`.").option("--auto-fix <mode>", "Auto-fix behaviour during script generation: 'interactive' (default, prompt y/N; declines on non-TTY), 'auto' (apply without prompt, for CI), 'skip' (agent-browser: apply only high-confidence fixes; external targets like playwright/runn: no fix pass at all).", parseAutoFixFlag, "interactive").option("--max-retries <n>", "Maximum number of auto-fix retries", "3").option("--force", "Overwrite previously generated test code without warning").option("--no-snapshot", "Don't pin AGENT_BROWSER_SESSION / capture page snapshots after a failure (debug toggle)").option("--update-agent-prompt", "After generation, ask Claude to refresh the target's \"<target>.agent\" learning prompt on the hub from a summary of the run. LLM-generating targets (playwright, runn) only; requires a hub connection.").option("--cwd <path>", "Working directory containing the .ccqa/ tree (monorepo support). Defaults to the current directory.").option("--project <name>", "Project name for the hub. Defaults to the current directory's name.")))).action(async (specPath, opts) => {
+const generateCommand = addHubOptions(addProfileOption(addLanguageOption(new Command("generate").argument("<feature/spec>", "Spec id in '<feature>/<spec>' form (resolves to .ccqa/features/<feature>/test-cases/<spec>/)").description("Generate test code from a spec via its target plugin. Recording-backed targets compile the existing ir.json (run `ccqa record` first); spec-input targets generate directly from the spec.").option("-m, --model <name>", "Claude model alias ('sonnet'|'opus'|'haiku') or full ID. Overrides CCQA_MODEL.").option("--target <id>", "Generate through this target instead of the spec's own — e.g. emit a Playwright spec from an agent-browser recording. The spec's `target:` stays the default for `ccqa run`.").option("--auto-fix <mode>", "Auto-fix behaviour during script generation: 'interactive' (default, prompt y/N; declines on non-TTY), 'auto' (apply without prompt, for CI), 'skip' (agent-browser: apply only high-confidence fixes; external targets like playwright/runn: no fix pass at all).", parseAutoFixFlag, "interactive").option("--max-retries <n>", "Maximum number of auto-fix retries", "3").option("--force", "Overwrite previously generated test code without warning").option("--no-snapshot", "Don't pin AGENT_BROWSER_SESSION / capture page snapshots after a failure (debug toggle)").option("--update-agent-prompt", "After generation, ask Claude to refresh the target's \"<target>.agent\" learning prompt on the hub from a summary of the run. LLM-generating targets (playwright, runn) only; requires a hub connection.").option("--cwd <path>", "Working directory containing the .ccqa/ tree (monorepo support). Defaults to the current directory.").option("--project <name>", "Project name for the hub. Defaults to the current directory's name.")))).action(withUsageErrors(async (specPath, opts) => {
 	const { featureName, specName } = parseSpecPath(specPath);
 	const language = opts.language ?? "auto";
 	const cwd = resolveCwd(opts.cwd);
@@ -15009,14 +15057,14 @@ const generateCommand = addHubOptions(addProfileOption(addLanguageOption(new Com
 		...language ? { language } : {},
 		...opts.model ? { model: opts.model } : {}
 	});
-});
+}));
 //#endregion
 //#region src/cli/record.ts
 const VALIDATION_MODES = ["lenient", "strict"];
 const recordCommand = addHubOptions(addProfileOption(addLanguageOption(new Command("record").argument("<feature/spec>", "Spec id in '<feature>/<spec>' form (resolves to .ccqa/features/<feature>/test-cases/<spec>/)").description("Record a test from a spec: run agent-browser to collect actions (trace), then compile them into runnable code via the spec's target (generate) — a vitest test.spec.ts for agent-browser, a @playwright/test spec for the playwright target. Recording-backed targets only; spec-input targets like runn have no trace step (use `ccqa generate`), and agent-browser live specs need no recording.").option("-m, --model <name>", "Claude model alias ('sonnet'|'opus'|'haiku') or full ID. Overrides CCQA_MODEL.").option("--validation-mode <mode>", "Post-trace validation behaviour: 'lenient' (default) tags failing actions; 'strict' drops them.", (raw) => {
 	if (VALIDATION_MODES.includes(raw)) return raw;
 	throw new Error(`--validation-mode must be one of ${VALIDATION_MODES.join(" | ")}`);
-}, "lenient").option("--auto-fix <mode>", "Auto-fix behaviour during script generation: 'interactive' (default, prompt y/N; declines on non-TTY), 'auto' (apply without prompt, for CI), 'skip' (agent-browser: apply only high-confidence fixes; external targets like playwright/runn: no fix pass at all).", parseAutoFixFlag, "interactive").option("--max-retries <n>", "Maximum number of auto-fix retries", "3").option("--force", "Overwrite an existing test.spec.ts without warning").option("--no-snapshot", "Don't pin AGENT_BROWSER_SESSION / capture page snapshots after a failure (debug toggle)").option("--skip-trace", "Skip the trace step and run codegen against an existing ir.json").option("--skip-codegen", "Run only the trace step (do not generate test.spec.ts)").option("--update-agent-prompt", "After the trace finishes, ask Claude to refresh the \"record.agent\" prompt on the hub from a summary of the run. Requires a hub connection.").option("--cwd <path>", "Working directory containing the .ccqa/ tree (monorepo support). Defaults to the current directory.").option("--project <name>", "Project name for the hub. Defaults to the current directory's name.")))).action(async (specPath, opts) => {
+}, "lenient").option("--auto-fix <mode>", "Auto-fix behaviour during script generation: 'interactive' (default, prompt y/N; declines on non-TTY), 'auto' (apply without prompt, for CI), 'skip' (agent-browser: apply only high-confidence fixes; external targets like playwright/runn: no fix pass at all).", parseAutoFixFlag, "interactive").option("--max-retries <n>", "Maximum number of auto-fix retries", "3").option("--force", "Overwrite an existing test.spec.ts without warning").option("--no-snapshot", "Don't pin AGENT_BROWSER_SESSION / capture page snapshots after a failure (debug toggle)").option("--skip-trace", "Skip the trace step and run codegen against an existing ir.json").option("--skip-codegen", "Run only the trace step (do not generate test.spec.ts)").option("--update-agent-prompt", "After the trace finishes, ask Claude to refresh the \"record.agent\" prompt on the hub from a summary of the run. Requires a hub connection.").option("--cwd <path>", "Working directory containing the .ccqa/ tree (monorepo support). Defaults to the current directory.").option("--project <name>", "Project name for the hub. Defaults to the current directory's name.")))).action(withUsageErrors(async (specPath, opts) => {
 	const { featureName, specName } = parseSpecPath(specPath);
 	const language = opts.language ?? "auto";
 	if (opts.skipTrace && opts.skipCodegen) {
@@ -15103,7 +15151,7 @@ const recordCommand = addHubOptions(addProfileOption(addLanguageOption(new Comma
 		...language ? { language } : {},
 		...opts.model ? { model: opts.model } : {}
 	});
-});
+}));
 /**
 * Compact summary of the trace pass for the record agent-prompt refresh.
 * Steps are reconstructed from the trace's status-line protocol (STEP_START
@@ -15416,34 +15464,6 @@ function driftResultsToReport(results, meta) {
 	};
 }
 //#endregion
-//#region src/cli/usage-errors.ts
-/**
-* Turn a `RunUsageError` into `[error] <message>` plus its exit code, for a
-* commander action.
-*
-* The helpers shared across commands — `resolveAnalysisBase`,
-* `collectChangedSpecs` — signal a bad invocation this way, so every command
-* that calls one needs the same boundary. Without it the process dies on an
-* unhandled rejection and prints a stack trace: `bin/ccqa.ts` installs no
-* global handler, so there is nowhere else for it to land.
-*
-* Lives here rather than beside `RunUsageError` because `src/run/errors.ts` is
-* deliberately dependency-free, and this needs the logger.
-*/
-function withUsageErrors(fn) {
-	return async (...args) => {
-		try {
-			await fn(...args);
-		} catch (err) {
-			if (err instanceof RunUsageError) {
-				error(err.message);
-				process.exit(err.exitCode);
-			}
-			throw err;
-		}
-	};
-}
-//#endregion
 //#region src/cli/drift.ts
 const DEFAULT_CONCURRENCY = 3;
 const driftCommand = addLanguageOption(new Command("drift").argument("[feature/spec]", "Optional spec id. If omitted, every spec under .ccqa/features/ is checked.").description("Standalone spec ↔ codebase static audit. Use for PR checks where the browser isn't run. For run-time audit with a structured report, see `ccqa run --report`.").option("--format <fmt>", "Output format: text | json | github", "text").option("--severity <level>", "Exit non-zero on this severity or higher: warn | error", "error").option("--concurrency <n>", `Parallel spec checks (default: ${DEFAULT_CONCURRENCY})`).option("-m, --model <name>", "Claude model alias ('sonnet'|'opus'|'haiku') or full ID. Overrides CCQA_MODEL.").option("--cwd <path>", "Working directory used as both the .ccqa root and the codebase Claude reads. Useful for monorepos. Defaults to process.cwd().").option("--changed", "Restrict drift checks to the specs a change reaches, decided by `ccqa select-specs` against --base (or, in CI, $GITHUB_BASE_REF). Costs one model call; specs it cannot decide are checked rather than skipped.").option("--base <ref>", "Base ref to diff against when --changed is set. Defaults to $GITHUB_BASE_REF (CI pull_request runs); required otherwise.").option("--push", "Push the drift result to a ccqa hub as a run (kind: drift).").option("--project <name>", "Logical project name for the pushed run. Defaults to the current directory's name.").option(...hubUrlOption).option(...hubTokenOption).option(...hubHeaderOption)).action(withUsageErrors(async (specPath, opts) => {

package/dist/package.json

@@ -1,6 +1,6 @@
 {
   "name": "ccqa",
-  "version": "1.10.0",
+  "version": "1.10.1",
   "type": "module",
   "description": "Browser test recorder powered by Claude Code and agent-browser",
   "repository": {

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "ccqa",
-  "version": "1.10.0",
+  "version": "1.10.1",
   "type": "module",
   "description": "Browser test recorder powered by Claude Code and agent-browser",
   "repository": {