feat(omo-claude): vendor start-work-continuation with claude session prefix

codex:->claude: session prefix; model/turn_id optional in Stop validator. QA: no-op without boulder.json, no crash. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-05-29 13:23:41 +09:00
parent 743671b43f
commit c284d8cf6b
11 changed files with 487 additions and 0 deletions
@@ -0,0 +1,50 @@
+<start-work-continuation>
+
+You are mid-flight on a Prometheus work plan. The turn just ended without finishing the plan. This is an automatic continuation — keep going. Do NOT ask the user whether to continue; the contract is auto-continue until every top-level checkbox is `- [x]`.
+
+# State
+
+- Plan: `{{PLAN_NAME}}`
+- Plan file: `{{PLAN_PATH}}`
+- Boulder state: `{{BOULDER_PATH}}`
+- Remaining top-level checkboxes: `{{REMAINING_COUNT}}` of `{{TOTAL_COUNT}}`
+- Next incomplete task: `{{NEXT_TASK_LABEL}}`
+{{WORKTREE_BLOCK}}
+- Ledger: `{{LEDGER_PATH}}`
+- Your session id in boulder.json: `codex:{{SESSION_ID}}`
+
+# What to do this turn
+
+1. Read `{{PLAN_PATH}}` AND `{{LEDGER_PATH}}` first — ground truth for what remains and what evidence has already been recorded. The plan checkbox and the ledger are the only sources of truth; do not trust your own memory of prior turns.
+2. Pick the FIRST unchecked top-level checkbox in `## TODOs` or `## Final Verification Wave`. Ignore nested checkboxes under Acceptance Criteria / Evidence / Definition of Done.
+3. Follow the `start-work` skill in full. The skill is already loaded from your earlier turn — re-read its file at `packages/omo-codex/plugin/skills/start-work/SKILL.md` if you have lost context.
+4. Decompose the checkbox into atomic sub-tasks. Dispatch them in PARALLEL via `spawn_agent` calls in this same response unless a sub-task has a NAMED blocking dependency (input from another sub-task or shared file).
+5. Every sub-task message MUST include all 6 sections and name one Manual-QA channel (HTTP call / tmux / browser use / computer use) with a captured artifact + a cleanup receipt. Tests are the floor; the channel artifact is the ceiling. Both are required.
+6. After verification of ALL sub-tasks under this checkbox: `apply_patch` the plan to change `- [ ]` → `- [x]`, re-read the plan to confirm the count decreased, append a `task-completed` line to the ledger, then continue.
+7. Do not start fresh on a sub-agent failure. Re-dispatch the same `task_name` with a fix-message: `FAILED: <exact error>` + `Diagnosis: <observation>` + `Fix: <instruction>`.
+
+# Hard constraints
+
+- No production code before a failing test exists. RED → GREEN → SURFACE.
+- No `--dry-run` as evidence. No "should work". No "tests pass" as completion proof.
+- No `as any` / `@ts-ignore` / `@ts-expect-error`. No deleting failing tests.
+- Cleanup receipt is mandatory. Leftover PIDs / `tmux` sessions / browser contexts / bound ports / containers / temp dirs = BLOCKED, not PASS.
+- The worktree path (if set in boulder.json) governs every file edit and command. Do not stray into the main repo.
+- session_ids you write to boulder.json MUST be prefixed `codex:`. Bare ids on read are legacy `opencode:`.
+
+# Stop conditions for THIS turn
+
+- A top-level checkbox flipped to `- [x]` after the 4-phase QA gate (Phase 1 read, Phase 2 automated, Phase 3 channel scenario, Phase 4 gate decision). Then the Stop hook will re-evaluate; if more checkboxes remain you will be continued again.
+- 3 same-failure cycles on one sub-task → escalate via `spawn_agent(agent_type="codex-ultrawork-reviewer", ...)` and stop dispatch.
+- Safety boundary (destructive command, secret exfiltration, production write) → stop and surface a safe substitute.
+- All top-level checkboxes `- [x]` AND (if gate triggered) `codex-ultrawork-reviewer` approved unconditionally → print the ORCHESTRATION COMPLETE block and end.
+
+# Output discipline
+
+- Surface only state changes: sub-agent dispatched, channel scenario PASS/FAIL with artifact path, checkbox marked, evidence appended to ledger.
+- Do NOT print "Should I continue?" — the Stop hook handles continuation.
+- Do NOT restate the full plan. Do NOT recap prior turns. The ledger and the plan file are the durable record.
+
+Begin now. Pick the next checkbox, dispatch the parallel sub-agents, verify, mark, continue.
+
+</start-work-continuation>
@@ -0,0 +1,26 @@
+{
+	"hooks": {
+		"Stop": [
+			{
+				"hooks": [
+					{
+						"type": "command",
+						"command": "node \"${CLAUDE_PLUGIN_ROOT}/components/start-work-continuation/dist/cli.js\" hook stop",
+						"timeout": 10
+					}
+				]
+			}
+		],
+		"SubagentStop": [
+			{
+				"hooks": [
+					{
+						"type": "command",
+						"command": "node \"${CLAUDE_PLUGIN_ROOT}/components/start-work-continuation/dist/cli.js\" hook subagent-stop",
+						"timeout": 10
+					}
+				]
+			}
+		]
+	}
+}
@@ -0,0 +1,53 @@
+{
+	"name": "@code-yeongyu/codex-start-work-continuation",
+	"version": "0.1.0",
+	"description": "Codex Stop hook continuation injector for omo-codex start-work plans.",
+	"type": "module",
+	"packageManager": "npm@11.12.1",
+	"license": "MIT",
+	"homepage": "https://github.com/code-yeongyu/codex-start-work-continuation",
+	"repository": {
+		"type": "git",
+		"url": "git+https://github.com/code-yeongyu/codex-start-work-continuation.git"
+	},
+	"bugs": {
+		"url": "https://github.com/code-yeongyu/codex-start-work-continuation/issues"
+	},
+	"keywords": [
+		"codex",
+		"codex-plugin",
+		"start-work",
+		"continuation",
+		"hooks",
+		"boulder"
+	],
+	"bin": {
+		"codex-start-work-continuation": "./dist/cli.js"
+	},
+	"scripts": {
+		"build": "tsc -p tsconfig.build.json",
+		"test": "vitest --run",
+		"test:watch": "vitest",
+		"typecheck": "tsc --noEmit",
+		"lint": "biome check .",
+		"lint:fix": "biome check --write .",
+		"check": "tsc --noEmit && biome check . && npm run build"
+	},
+	"files": [
+		"dist",
+		"directive.md",
+		"hooks",
+		"README.md",
+		"LICENSE",
+		"NOTICE"
+	],
+	"devDependencies": {
+		"@biomejs/biome": "2.4.15",
+		"@types/node": "^25.7.0",
+		"typescript": "^6.0.3",
+		"vitest": "^4.1.5"
+	},
+	"engines": {
+		"node": ">=20.0.0"
+	}
+}
@@ -0,0 +1,167 @@
+import { isAbsolute, join, resolve } from "node:path";
+
+import type { ReadonlyFileSystem } from "./types.js";
+
+const CHECKBOX_PATTERN = /^- \[[ xX]\] /;
+const UNCHECKED_PATTERN = /^- \[ \] /;
+const TODO_HEADING = "TODOs";
+const FINAL_VERIFICATION_HEADING = "Final Verification Wave";
+
+type WorkStatus = "active" | "completed" | "paused" | "abandoned";
+
+type BoulderWork = {
+	readonly activePlan: string;
+	readonly planName: string;
+	readonly status: WorkStatus;
+	readonly sessionIds: readonly string[];
+	readonly worktreePath: string | null;
+};
+
+export type PlanChecklist = {
+	readonly remaining: number;
+	readonly total: number;
+	readonly nextTaskLabel: string | null;
+};
+
+export type ContinuationState = {
+	readonly planName: string;
+	readonly planPath: string;
+	readonly boulderPath: string;
+	readonly ledgerPath: string;
+	readonly worktreePath: string | null;
+	readonly checklist: PlanChecklist;
+};
+
+export function parsePlanChecklist(markdown: string): PlanChecklist {
+	const lines = markdown.split(/\r?\n/);
+	const hasCountedSections = lines.some(hasCountedSectionHeading);
+	let remaining = 0;
+	let total = 0;
+	let nextTaskLabel: string | null = null;
+	let isCountedSection = !hasCountedSections;
+	for (const line of lines) {
+		const heading = parseLevelTwoHeading(line);
+		if (heading !== null) isCountedSection = isCountedHeading(heading);
+		if (!isCountedSection) continue;
+		if (!CHECKBOX_PATTERN.test(line)) continue;
+		total += 1;
+		if (!UNCHECKED_PATTERN.test(line)) continue;
+		remaining += 1;
+		if (nextTaskLabel === null) nextTaskLabel = line.slice("- [ ] ".length);
+	}
+	return { remaining, total, nextTaskLabel };
+}
+
+function hasCountedSectionHeading(line: string): boolean {
+	const heading = parseLevelTwoHeading(line);
+	return heading !== null && isCountedHeading(heading);
+}
+
+export function readContinuationState(
+	cwd: string,
+	sessionId: string,
+	fs: ReadonlyFileSystem,
+): ContinuationState | null {
+	const boulderPath = join(cwd, ".omo", "boulder.json");
+	const boulderText = readTextFile(fs, boulderPath);
+	if (boulderText === null) return null;
+	const parsed = parseJsonObject(boulderText);
+	if (parsed === null) return null;
+	const work = findMatchingWork(parsed, `claude:${sessionId}`);
+	if (work === null) return null;
+	const planPath = resolvePlanPath(cwd, work.activePlan);
+	const planText = readTextFile(fs, planPath);
+	if (planText === null) return null;
+	const checklist = parsePlanChecklist(planText);
+	if (checklist.remaining === 0) return null;
+	return {
+		planName: work.planName,
+		planPath,
+		boulderPath,
+		ledgerPath: join(cwd, ".omo", "start-work", "ledger.jsonl"),
+		worktreePath: work.worktreePath,
+		checklist,
+	};
+}
+
+function findMatchingWork(state: Record<string, unknown>, prefixedSessionId: string): BoulderWork | null {
+	const worksValue = state["works"];
+	const candidates = isRecord(worksValue) ? Object.values(worksValue) : [state];
+	for (const candidate of candidates) {
+		const work = parseBoulderWork(candidate);
+		if (work === null) continue;
+		if (!isContinuableStatus(work.status)) continue;
+		if (work.sessionIds.includes(prefixedSessionId)) return work;
+	}
+	return null;
+}
+
+function parseBoulderWork(value: unknown): BoulderWork | null {
+	if (!isRecord(value)) return null;
+	const activePlan = value["active_plan"];
+	const planName = value["plan_name"];
+	const status = parseWorkStatus(value["status"]);
+	const sessionIds = value["session_ids"];
+	const worktreePath = value["worktree_path"];
+	if (typeof activePlan !== "string") return null;
+	if (typeof planName !== "string") return null;
+	if (status === null) return null;
+	if (!isStringArray(sessionIds)) return null;
+	return {
+		activePlan,
+		planName,
+		status,
+		sessionIds,
+		worktreePath: typeof worktreePath === "string" ? worktreePath : null,
+	};
+}
+
+function parseWorkStatus(value: unknown): WorkStatus | null {
+	if (value === "active" || value === "completed" || value === "paused" || value === "abandoned") return value;
+	return null;
+}
+
+function isContinuableStatus(status: WorkStatus): boolean {
+	return status === "active" || status === "paused";
+}
+
+function parseLevelTwoHeading(line: string): string | null {
+	if (!line.startsWith("## ")) return null;
+	if (line.startsWith("### ")) return null;
+	return line.slice("## ".length).trim();
+}
+
+function isCountedHeading(heading: string): boolean {
+	return heading === TODO_HEADING || heading === FINAL_VERIFICATION_HEADING;
+}
+
+function resolvePlanPath(cwd: string, activePlan: string): string {
+	return isAbsolute(activePlan) ? activePlan : resolve(cwd, activePlan);
+}
+
+function readTextFile(fs: ReadonlyFileSystem, path: string): string | null {
+	try {
+		return fs.readFileSync(path, "utf8");
+	} catch (error) {
+		if (error instanceof Error) return null;
+		throw error;
+	}
+}
+
+function parseJsonObject(json: string): Record<string, unknown> | null {
+	try {
+		const parsed: unknown = JSON.parse(json);
+		return isRecord(parsed) ? parsed : null;
+	} catch (error) {
+		if (error instanceof SyntaxError) return null;
+		throw error;
+	}
+}
+
+function isStringArray(value: unknown): value is readonly string[] {
+	return Array.isArray(value) && value.every((item) => typeof item === "string");
+}
+
+function isRecord(value: unknown): value is Record<string, unknown> {
+	return typeof value === "object" && value !== null && !Array.isArray(value);
+}
@@ -0,0 +1,52 @@
+#!/usr/bin/env node
+import { readFileSync } from "node:fs";
+import { stdin as processStdin, stdout as processStdout } from "node:process";
+
+import { runStopHook } from "./codex-hook.js";
+import type { ReadonlyFileSystem } from "./types.js";
+
+const nodeFileSystem: ReadonlyFileSystem = {
+	readFileSync(path, encoding) {
+		return readFileSync(path, encoding);
+	},
+};
+
+const command = process.argv[2];
+const subcommand = process.argv[3];
+
+if (command === "hook" && (subcommand === "stop" || subcommand === "subagent-stop")) {
+	await runHookCli();
+} else {
+	process.stderr.write("Usage: codex-start-work-continuation hook <stop|subagent-stop>\n");
+	process.exitCode = 1;
+}
+
+async function runHookCli(): Promise<void> {
+	const raw = await readStdin();
+	if (raw.trim().length === 0) return;
+	const parsed = parseHookInput(raw);
+	const output = runStopHook(parsed, nodeFileSystem);
+	if (output.length > 0) processStdout.write(output);
+}
+
+function parseHookInput(raw: string): unknown | undefined {
+	try {
+		const parsed: unknown = JSON.parse(raw);
+		return parsed;
+	} catch (error) {
+		if (error instanceof SyntaxError) return undefined;
+		throw error;
+	}
+}
+
+function readStdin(): Promise<string> {
+	return new Promise((resolve) => {
+		let data = "";
+		processStdin.setEncoding("utf8");
+		processStdin.on("data", (chunk: string) => {
+			data += chunk;
+		});
+		processStdin.once("error", () => resolve(data));
+		processStdin.once("end", () => resolve(data));
+	});
+}
@@ -0,0 +1,66 @@
+import type { ContinuationState } from "./boulder-reader.js";
+import { readContinuationState } from "./boulder-reader.js";
+import { START_WORK_CONTINUATION_DIRECTIVE } from "./directive.js";
+import type { ReadonlyFileSystem, StopHookEventName, StopHookOutput, StopInput } from "./types.js";
+
+export function runStopHook(input: unknown, fs: ReadonlyFileSystem): string {
+	if (!isStopInput(input)) return "";
+	if (input.stop_hook_active) return "";
+	const state = readContinuationState(input.cwd, input.session_id, fs);
+	if (state === null) return "";
+	return JSON.stringify({
+		decision: "block",
+		reason: renderDirective(state, input.session_id),
+	} satisfies StopHookOutput);
+}
+
+function renderDirective(state: ContinuationState, sessionId: string): string {
+	const lineBreak = String.fromCharCode(10);
+	const worktreeBlock =
+		state.worktreePath === null
+			? ""
+			: `${lineBreak}- Worktree: \`${state.worktreePath}\` (all edits, tests, and commands run inside this directory)`;
+	const replacements = {
+		PLAN_NAME: state.planName,
+		PLAN_PATH: state.planPath,
+		BOULDER_PATH: state.boulderPath,
+		REMAINING_COUNT: String(state.checklist.remaining),
+		TOTAL_COUNT: String(state.checklist.total),
+		NEXT_TASK_LABEL: state.checklist.nextTaskLabel ?? "",
+		WORKTREE_BLOCK: worktreeBlock,
+		LEDGER_PATH: state.ledgerPath,
+		SESSION_ID: sessionId,
+	} as const;
+	let rendered = START_WORK_CONTINUATION_DIRECTIVE;
+	for (const [placeholder, value] of Object.entries(replacements)) {
+		rendered = rendered.replaceAll(`{{${placeholder}}}`, value);
+	}
+	return rendered;
+}
+
+function isStopInput(value: unknown): value is StopInput {
+	return (
+		isRecord(value) &&
+		isStopHookEventName(value["hook_event_name"]) &&
+		typeof value["session_id"] === "string" &&
+		(typeof value["turn_id"] === "string" || value["turn_id"] === undefined) &&
+		typeof value["transcript_path"] === "string" &&
+		typeof value["cwd"] === "string" &&
+		(typeof value["model"] === "string" || value["model"] === undefined) &&
+		typeof value["permission_mode"] === "string" &&
+		typeof value["stop_hook_active"] === "boolean" &&
+		optionalString(value["last_assistant_message"])
+	);
+}
+
+function isStopHookEventName(value: unknown): value is StopHookEventName {
+	return value === "Stop" || value === "SubagentStop";
+}
+
+function optionalString(value: unknown): boolean {
+	return value === undefined || typeof value === "string";
+}
+
+function isRecord(value: unknown): value is Record<string, unknown> {
+	return typeof value === "object" && value !== null && !Array.isArray(value);
+}
@@ -0,0 +1,6 @@
+import { readFileSync } from "node:fs";
+
+export const START_WORK_CONTINUATION_DIRECTIVE: string = readFileSync(
+	new URL("../directive.md", import.meta.url),
+	"utf8",
+);
@@ -0,0 +1,5 @@
+export type { ContinuationState, PlanChecklist } from "./boulder-reader.js";
+export { parsePlanChecklist, readContinuationState } from "./boulder-reader.js";
+export { runStopHook } from "./codex-hook.js";
+export { START_WORK_CONTINUATION_DIRECTIVE } from "./directive.js";
+export type { ReadonlyFileSystem, StopHookEventName, StopHookOutput, StopInput } from "./types.js";
@@ -0,0 +1,23 @@
+export const STOP_HOOK_EVENTS = ["Stop", "SubagentStop"] as const;
+export type StopHookEventName = (typeof STOP_HOOK_EVENTS)[number];
+
+export type StopInput = {
+	readonly hook_event_name: StopHookEventName;
+	readonly session_id: string;
+	readonly turn_id: string;
+	readonly transcript_path: string;
+	readonly cwd: string;
+	readonly model: string;
+	readonly permission_mode: string;
+	readonly stop_hook_active: boolean;
+	readonly last_assistant_message?: string;
+};
+
+export type StopHookOutput = {
+	readonly decision: "block";
+	readonly reason: string;
+};
+
+export type ReadonlyFileSystem = {
+	readFileSync(path: string, encoding: "utf8"): string;
+};
@@ -0,0 +1,12 @@
+{
+	"extends": "./tsconfig.json",
+	"compilerOptions": {
+		"allowImportingTsExtensions": false,
+		"declaration": true,
+		"outDir": "dist",
+		"rootDir": "src",
+		"noEmit": false
+	},
+	"include": ["src/**/*"],
+	"exclude": ["test/**/*"]
+}
@@ -0,0 +1,27 @@
+{
+	"compilerOptions": {
+		"target": "ESNext",
+		"module": "NodeNext",
+		"moduleResolution": "NodeNext",
+		"lib": ["ESNext"],
+		"strict": true,
+		"exactOptionalPropertyTypes": true,
+		"noUncheckedIndexedAccess": true,
+		"noPropertyAccessFromIndexSignature": true,
+		"verbatimModuleSyntax": true,
+		"noImplicitOverride": true,
+		"noImplicitReturns": true,
+		"noFallthroughCasesInSwitch": true,
+		"noUnusedLocals": true,
+		"noUnusedParameters": true,
+		"esModuleInterop": true,
+		"allowImportingTsExtensions": true,
+		"skipLibCheck": true,
+		"forceConsistentCasingInFileNames": true,
+		"resolveJsonModule": true,
+		"useDefineForClassFields": false,
+		"types": ["node"],
+		"noEmit": true
+	},
+	"include": ["src/**/*", "test/**/*", "vitest.config.ts"]
+}