@oh-my-pi/pi-coding-agent 11.8.0 → 11.8.2

package/CHANGELOG.md

@@ -2,6 +2,31 @@
 
 ## [Unreleased]
 
+## [11.8.1] - 2026-02-10
+
+### Added
+
+- Added current date to system prompt context in YYYY-MM-DD format for date-aware agent reasoning
+- Added file size display in UI when files are skipped due to size limits
+- Added support for gigabyte (GB) file size formatting in truncate utility
+
+### Changed
+
+- Changed skipped file messages to include file size information for better visibility into why files were excluded
+- Changed file processing to skip reading files exceeding 5MB (text) or 25MB (images) and include them as path-only references instead
+- Changed @mention auto-reading to skip files exceeding 5MB (text) or 25MB (images) to prevent out-of-memory issues with large files
+- Clarified that subagents automatically inherit full system prompt including AGENTS.md, context files, and skills — do not repeat project rules or conventions in task context
+- Updated task context guidance to focus on session-specific information subagents lack, eliminating redundant documentation of project constraints already available to them
+- Refined constraints template to emphasize task-specific rules and session decisions rather than global project conventions
+- Expanded anti-patterns section to explicitly flag redundant context that wastes tokens by repeating AGENTS.md rules, project constraints, and tool preferences
+
+### Fixed
+
+- Fixed bash tool hanging when commands spawn background jobs by properly detecting foreground process completion
+- Fixed bash tool occasionally hanging after command completion when background jobs keep stdout/stderr open
+- Fixed crash when auto-reading @mentions for very large files by skipping content injection with an explicit "skipped" note
+- Improved bash tool output draining after foreground completion to reduce tail output truncation
+
 ## [11.8.0] - 2026-02-10
 ### Added
 

package/package.json

@@ -1,6 +1,6 @@
 {
 	"name": "@oh-my-pi/pi-coding-agent",
-	"version": "11.8.0",
+	"version": "11.8.2",
 	"description": "Coding agent CLI with read, bash, edit, write tools and session management",
 	"type": "module",
 	"ompConfig": {
@@ -90,12 +90,12 @@
 		"@mozilla/readability": "0.6.0",
 		"@oclif/core": "^4.8.0",
 		"@oclif/plugin-autocomplete": "^3.2.40",
-		"@oh-my-pi/omp-stats": "11.8.0",
-		"@oh-my-pi/pi-agent-core": "11.8.0",
-		"@oh-my-pi/pi-ai": "11.8.0",
-		"@oh-my-pi/pi-natives": "11.8.0",
-		"@oh-my-pi/pi-tui": "11.8.0",
-		"@oh-my-pi/pi-utils": "11.8.0",
+		"@oh-my-pi/omp-stats": "11.8.2",
+		"@oh-my-pi/pi-agent-core": "11.8.2",
+		"@oh-my-pi/pi-ai": "11.8.2",
+		"@oh-my-pi/pi-natives": "11.8.2",
+		"@oh-my-pi/pi-tui": "11.8.2",
+		"@oh-my-pi/pi-utils": "11.8.2",
 		"@sinclair/typebox": "^0.34.48",
 		"ajv": "^8.17.1",
 		"chalk": "^5.6.2",

package/src/cli/file-processor.ts

@@ -7,9 +7,15 @@ import type { ImageContent } from "@oh-my-pi/pi-ai";
 import { isEnoent } from "@oh-my-pi/pi-utils";
 import chalk from "chalk";
 import { resolveReadPath } from "../tools/path-utils";
+import { formatSize } from "../tools/truncate";
 import { formatDimensionNote, resizeImage } from "../utils/image-resize";
 import { detectSupportedImageMimeTypeFromFile } from "../utils/mime";
 
+// Keep CLI startup responsive and avoid OOM when users pass huge files.
+// If a file exceeds these limits, we include it as a path-only <file/> block.
+const MAX_CLI_TEXT_BYTES = 5 * 1024 * 1024; // 5MB
+const MAX_CLI_IMAGE_BYTES = 25 * 1024 * 1024; // 25MB
+
 export interface ProcessedFiles {
 	text: string;
 	images: ImageContent[];
@@ -30,6 +36,27 @@ export async function processFileArguments(fileArgs: string[], options?: Process
 		// Expand and resolve path (handles ~ expansion and macOS screenshot Unicode spaces)
 		const absolutePath = path.resolve(resolveReadPath(fileArg, process.cwd()));
 
+		let stat: Awaited<ReturnType<typeof fs.stat>>;
+		try {
+			stat = await fs.stat(absolutePath);
+		} catch (err) {
+			if (isEnoent(err)) {
+				console.error(chalk.red(`Error: File not found: ${absolutePath}`));
+				process.exit(1);
+			}
+			throw err;
+		}
+
+		const mimeType = await detectSupportedImageMimeTypeFromFile(absolutePath);
+		const maxBytes = mimeType ? MAX_CLI_IMAGE_BYTES : MAX_CLI_TEXT_BYTES;
+		if (stat.size > maxBytes) {
+			console.error(
+				chalk.yellow(`Warning: Skipping file contents (too large: ${formatSize(stat.size)}): ${absolutePath}`),
+			);
+			text += `<file name="${absolutePath}">(skipped: too large, ${formatSize(stat.size)})</file>\n`;
+			continue;
+		}
+
 		// Read file, handling not-found gracefully
 		let buffer: Buffer;
 		try {
@@ -45,8 +72,6 @@ export async function processFileArguments(fileArgs: string[], options?: Process
 			continue;
 		}
 
-		const mimeType = await detectSupportedImageMimeTypeFromFile(absolutePath);
-
 		if (mimeType) {
 			// Handle image file
 			const base64Content = buffer.toBase64();

package/src/modes/utils/ui-helpers.ts

@@ -17,6 +17,7 @@ import { theme } from "../../modes/theme/theme";
 import type { CompactionQueuedMessage, InteractiveModeContext } from "../../modes/types";
 import { type CustomMessage, SKILL_PROMPT_MESSAGE_TYPE, type SkillPromptDetails } from "../../session/messages";
 import type { SessionContext } from "../../session/session-manager";
+import { formatSize } from "../../tools/truncate";
 
 type TextBlock = { type: "text"; text: string };
 
@@ -127,11 +128,17 @@ export class UiHelpers {
 			case "fileMention": {
 				// Render compact file mention display
 				for (const file of message.files) {
-					const suffix = file.image
-						? "(image)"
-						: file.lineCount === undefined
-							? "(unknown lines)"
-							: `(${file.lineCount} lines)`;
+					let suffix: string;
+					if (file.skippedReason === "tooLarge") {
+						const size = typeof file.byteSize === "number" ? formatSize(file.byteSize) : "unknown size";
+						suffix = `(skipped: ${size})`;
+					} else {
+						suffix = file.image
+							? "(image)"
+							: file.lineCount === undefined
+								? "(unknown lines)"
+								: `(${file.lineCount} lines)`;
+					}
 					const text = `${theme.fg("dim", `${theme.tree.last} `)}${theme.fg("muted", "Read")} ${theme.fg(
 						"accent",
 						file.path,

package/src/prompts/system/system-prompt.md

@@ -255,6 +255,7 @@ Read `rule://<name>` when working in domain.
 {{/if}}
 
 Current directory: {{cwd}}
+Current date: {{date}}
 
 <north_star>
 Correctness. Usefulness. Fidelity to truth.

package/src/prompts/tools/task.md

@@ -1,13 +1,18 @@
 # Task
 
-Launch subagents to execute parallel, well-scoped tasks with shared, token-efficient context.
+Launch subagents to execute parallel, well-scoped tasks.
 
-Subagents have **zero implicit context** — they see only `context` + `assignment`. Treat every subagent as a senior engineer on day one: technically strong, but unfamiliar with every decision, convention, and file layout you've accumulated.
+## What subagents inherit automatically
+Subagents receive the **full system prompt**, including AGENTS.md, context files, and skills. Do NOT repeat project rules, coding conventions, or style guidelines in `context` — they already have them.
 
-Subagents CAN grep the parent conversation file for supplementary details, but CANNOT grep for:
+## What subagents do NOT have
+Subagents have no access to your conversation history. They don't know:
 - Decisions you made but didn't write down
-- Conventions that exist only in your head
-- Which of 50 possible approaches you want
+- Which approach you chose among alternatives
+- What you learned from reading files during this session
+- Requirements the user stated only in conversation
+
+Subagents CAN grep the parent conversation file for supplementary details.
 ---
 
 ## Parameters
@@ -18,9 +23,13 @@ Agent type for all tasks in this batch.
 
 ### `context` (optional — strongly recommended)
 
-Shared background prepended verbatim to every task `assignment`. Common info once; reduces token cost.
+Shared background prepended verbatim to every task `assignment`. Use only for session-specific information subagents lack.
+
+<critical>
+Do NOT include project rules, coding conventions, or style guidelines — subagents already have AGENTS.md and context files in their system prompt. Repeating them wastes tokens.
+</critical>
 
-Use template; omit non-applicable sections.
+Use template; omit non-applicable sections:
 
 ````
 ## Goal
@@ -30,9 +39,8 @@ One sentence: batch accomplishes together.
 Explicitly exclude tempting scope — what tasks must not touch/attempt.
 
 ## Constraints
-- MUST / MUST NOT rules (naming, error handling, banned approaches)
-- Language/framework version requirements
-- What exists vs what to create
+- Task-specific MUST / MUST NOT rules not already in AGENTS.md
+- Decisions made during this session that affect implementation
 
 ## Reference Files
 - `path/to/file.ext` — pattern demo
@@ -47,9 +55,9 @@ Explicitly exclude tempting scope — what tasks must not touch/attempt.
 - Definition of "done" for batch
 - Note: build/test/lint verification happens AFTER all tasks complete — not inside tasks (see below)
 ````
-**Belongs in `context`**: project goal, non-goals, constraints, conventions, reference paths, shared type definitions, API contracts, global acceptance commands — anything 2+ tasks need.
+**Belongs in `context`**: task-specific goal, non-goals, session decisions, reference paths, shared type definitions, API contracts, global acceptance commands — anything 2+ tasks need that isn't already in AGENTS.md.
 **Rule of thumb:** if repeat in 2+ tasks, belongs in `context`.
-**Does NOT belong in `context`**: per-task file lists, one-off requirements (go in `assignment`), structured output format (goes in `schema`).
+**Does NOT belong in `context`**: project rules already in AGENTS.md/context files, per-task file lists, one-off requirements (go in `assignment`), structured output format (goes in `schema`).
 
 ### `tasks` (required)
 
@@ -117,6 +125,10 @@ Use structure every assignment:
 - "Use existing patterns."
 - "Follow conventions."
 - "No WASM."
+**Redundant context** — wastes tokens repeating what subagents already have:
+- Restating AGENTS.md rules (coding style, import conventions, logger usage)
+- Repeating project constraints from context files
+- Listing tool/framework preferences already documented in the repo
 
 If tempted to write above, expand using templates.
 **Output format in prose instead of `schema`** — agent returns null:
@@ -252,7 +264,7 @@ Layered work with dependencies:
 ## Pre-flight checklist
 
 Before calling tool, verify:
-- [ ] `context` includes shared constraints, references, definition of done
+- [ ] `context` includes only session-specific info not already in AGENTS.md/context files
 - [ ] Each `assignment` follows assignment template — not one-liner
 - [ ] Each `assignment` includes edge cases / "don’t break" items
 - [ ] Tasks truly parallel (no hidden dependencies)

package/src/session/messages.ts

@@ -114,6 +114,10 @@ export interface FileMentionMessage {
 		path: string;
 		content: string;
 		lineCount?: number;
+		/** File size in bytes, if known. */
+		byteSize?: number;
+		/** Why the file contents were omitted from auto-read. */
+		skippedReason?: "tooLarge";
 		image?: ImageContent;
 	}>;
 	timestamp: number;

package/src/system-prompt.ts

@@ -473,6 +473,11 @@ export async function buildSystemPrompt(options: BuildSystemPromptOptions = {}):
 	const systemPromptCustomization = await loadSystemPromptFiles({ cwd: resolvedCwd });
 
 	const now = new Date();
+	const date = now.toLocaleDateString("en-CA", {
+		year: "numeric",
+		month: "2-digit",
+		day: "2-digit",
+	});
 	const dateTime = now.toLocaleString("en-US", {
 		weekday: "long",
 		year: "numeric",
@@ -529,6 +534,7 @@ export async function buildSystemPrompt(options: BuildSystemPromptOptions = {}):
 			skills: filteredSkills,
 			preloadedSkills: preloadedSkillContents,
 			rules: rules ?? [],
+			date,
 			dateTime,
 			cwd: resolvedCwd,
 		});
@@ -544,6 +550,7 @@ export async function buildSystemPrompt(options: BuildSystemPromptOptions = {}):
 		skills: filteredSkills,
 		preloadedSkills: preloadedSkillContents,
 		rules: rules ?? [],
+		date,
 		dateTime,
 		cwd: resolvedCwd,
 		appendSystemPrompt: resolvedAppendPrompt ?? "",

package/src/tools/truncate.ts

@@ -53,8 +53,10 @@ export function formatSize(bytes: number): string {
 		return `${bytes}B`;
 	} else if (bytes < 1024 * 1024) {
 		return `${(bytes / 1024).toFixed(1)}KB`;
-	} else {
+	} else if (bytes < 1024 * 1024 * 1024) {
 		return `${(bytes / (1024 * 1024)).toFixed(1)}MB`;
+	} else {
+		return `${(bytes / (1024 * 1024 * 1024)).toFixed(1)}GB`;
 	}
 }
 

package/src/utils/file-mentions.ts

@@ -22,6 +22,11 @@ const TRAILING_PUNCTUATION_REGEX = /[)\]}>.,;:!?"'`]+$/;
 const MENTION_BOUNDARY_REGEX = /[\s([{<"'`]/;
 const DEFAULT_DIR_LIMIT = 500;
 
+// Avoid OOM when users @mention very large files. Above these limits we skip
+// auto-reading and only include the path in the message.
+const MAX_AUTO_READ_TEXT_BYTES = 5 * 1024 * 1024; // 5MB
+const MAX_AUTO_READ_IMAGE_BYTES = 25 * 1024 * 1024; // 25MB
+
 function isMentionBoundary(text: string, index: number): boolean {
 	if (index === 0) return true;
 	return MENTION_BOUNDARY_REGEX.test(text[index - 1]);
@@ -183,6 +188,15 @@ export async function generateFileMentionMessages(
 
 			const mimeType = await detectSupportedImageMimeTypeFromFile(absolutePath);
 			if (mimeType) {
+				if (stat.size > MAX_AUTO_READ_IMAGE_BYTES) {
+					files.push({
+						path: filePath,
+						content: `(skipped auto-read: too large, ${formatSize(stat.size)})`,
+						byteSize: stat.size,
+						skippedReason: "tooLarge",
+					});
+					continue;
+				}
 				const buffer = await fs.readFile(absolutePath);
 				if (buffer.length === 0) {
 					continue;
@@ -210,6 +224,16 @@ export async function generateFileMentionMessages(
 				continue;
 			}
 
+			if (stat.size > MAX_AUTO_READ_TEXT_BYTES) {
+				files.push({
+					path: filePath,
+					content: `(skipped auto-read: too large, ${formatSize(stat.size)})`,
+					byteSize: stat.size,
+					skippedReason: "tooLarge",
+				});
+				continue;
+			}
+
 			const content = await Bun.file(absolutePath).text();
 			const { output, lineCount } = buildTextOutput(content);
 			files.push({ path: filePath, content: output, lineCount });