@oh-my-pi/pi-coding-agent 13.9.6 → 13.9.11

package/CHANGELOG.md

@@ -2,6 +2,37 @@
 
 ## [Unreleased]
 
+## [13.9.10] - 2026-03-08
+### Added
+
+- Added `env` parameter to bash tool to pass environment variables safely without shell re-parsing, preventing quote and special character bugs with multiline or untrusted values
+- Added support for rendering partial `env` assignments in command preview while tool arguments are still streaming
+- Added `env` support to the bash tool so commands can reference safe shell variables without inline quoting bugs for multiline or quote-heavy values
+
+### Changed
+
+- Changed bash tool to display environment variable assignments in command preview when `env` parameter is used
+
+## [13.9.8] - 2026-03-08
+### Added
+
+- Added docs.rs scraper for extracting Rust crate documentation from rustdoc JSON, including support for modules, functions, structs, traits, enums, and other Rust items with caching
+
+## [13.9.7] - 2026-03-08
+
+### Added
+
+- Added `skipPostPromptRecoveryWait` option to handoff operations to defer recovery work until after handoff completion
+- Added deferred auto-compaction scheduling to allow threshold-triggered handoffs to complete while the original prompt is still unwinding
+
+### Changed
+
+- Extracted handoff document template to dedicated prompt file for improved maintainability and template variable support
+- Changed handoff prompt generation to use template rendering with support for custom focus instructions
+- Refactored internal prompt-in-flight tracking from boolean flag to counter to properly handle nested prompt operations
+- Moved llms.txt endpoint discovery to fallback strategy when rendered page content is low quality, prioritizing page-specific content over site-wide files
+- Enhanced llms.txt endpoint detection to scope candidates to the requested URL path, searching section-specific files before site-wide ones
+
 ## [13.9.6] - 2026-03-08
 
 ### Added

package/package.json

@@ -1,7 +1,7 @@
 {
 	"type": "module",
 	"name": "@oh-my-pi/pi-coding-agent",
-	"version": "13.9.6",
+	"version": "13.9.11",
 	"description": "Coding agent CLI with read, bash, edit, write tools and session management",
 	"homepage": "https://github.com/can1357/oh-my-pi",
 	"author": "Can Boluk",
@@ -41,12 +41,12 @@
 	},
 	"dependencies": {
 		"@mozilla/readability": "^0.6",
-		"@oh-my-pi/omp-stats": "13.9.6",
-		"@oh-my-pi/pi-agent-core": "13.9.6",
-		"@oh-my-pi/pi-ai": "13.9.6",
-		"@oh-my-pi/pi-natives": "13.9.6",
-		"@oh-my-pi/pi-tui": "13.9.6",
-		"@oh-my-pi/pi-utils": "13.9.6",
+		"@oh-my-pi/omp-stats": "13.9.11",
+		"@oh-my-pi/pi-agent-core": "13.9.11",
+		"@oh-my-pi/pi-ai": "13.9.11",
+		"@oh-my-pi/pi-natives": "13.9.11",
+		"@oh-my-pi/pi-tui": "13.9.11",
+		"@oh-my-pi/pi-utils": "13.9.11",
 		"@sinclair/typebox": "^0.34",
 		"@xterm/headless": "^6.0",
 		"ajv": "^8.18",

package/src/modes/components/tool-execution.ts

@@ -478,7 +478,7 @@ export class ToolExecutionComponent extends Container {
 						},
 						this.#renderState,
 						theme,
-						this.#args, // Pass args for tools that need them
+						this.#getCallArgsForRender(),
 					);
 					if (resultComponent) {
 						this.#contentBox.addChild(ensureInvalidate(resultComponent));
@@ -560,10 +560,14 @@ export class ToolExecutionComponent extends Container {
 			return Math.max(1, Math.min(maxSeconds, value));
 		};
 
-		if (this.#toolName === "bash" && this.#result) {
-			// Pass raw output and expanded state - renderer handles width-aware truncation
-			const output = this.#getTextOutput().trimEnd();
-			context.output = output;
+		if (this.#toolName === "bash") {
+			// Bash needs render context even before a result exists. The renderer uses the pending-call args
+			// plus this context to keep the inline command preview visible while tool-call JSON is still streaming.
+			if (this.#result) {
+				// Pass raw output and expanded state - renderer handles width-aware truncation
+				const output = this.#getTextOutput().trimEnd();
+				context.output = output;
+			}
 			context.expanded = this.#expanded;
 			context.previewLines = BASH_DEFAULT_PREVIEW_LINES;
 			context.timeout = normalizeTimeoutSeconds(this.#args?.timeout, 3600);

package/src/modes/controllers/event-controller.ts

@@ -184,13 +184,19 @@ export class EventController {
 							continue;
 						}
 
+						// Preserve the raw partial JSON for renderers that need to surface fields before the JSON object closes.
+						// Bash uses this to show inline env assignments during streaming instead of popping them in at completion.
+						const renderArgs =
+							"partialJson" in content
+								? { ...content.arguments, __partialJson: content.partialJson }
+								: content.arguments;
 						if (!this.ctx.pendingTools.has(content.id)) {
 							this.#resetReadGroup();
 							this.ctx.chatContainer.addChild(new Text("", 0, 0));
 							const tool = this.ctx.session.getToolByName(content.name);
 							const component = new ToolExecutionComponent(
 								content.name,
-								content.arguments,
+								renderArgs,
 								{
 									showImages: settings.get("terminal.showImages"),
 									editFuzzyThreshold: settings.get("edit.fuzzyThreshold"),
@@ -206,7 +212,7 @@ export class EventController {
 						} else {
 							const component = this.ctx.pendingTools.get(content.id);
 							if (component) {
-								component.updateArgs(content.arguments, content.id);
+								component.updateArgs(renderArgs, content.id);
 							}
 						}
 					}

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

@@ -272,9 +272,13 @@ export class UiHelpers {
 
 					readGroup = null;
 					const tool = this.ctx.session.getToolByName(content.name);
+					const renderArgs =
+						"partialJson" in content
+							? { ...content.arguments, __partialJson: content.partialJson }
+							: content.arguments;
 					const component = new ToolExecutionComponent(
 						content.name,
-						content.arguments,
+						renderArgs,
 						{
 							showImages: settings.get("terminal.showImages"),
 							editFuzzyThreshold: settings.get("edit.fuzzyThreshold"),

package/src/prompts/agents/explore.md

@@ -99,6 +99,7 @@ Given a task, you rapidly investigate the codebase and return structured finding
 <directives>
 - You **MUST** use tools for broad pattern matching / code search as much as possible.
 - You **SHOULD** invoke tools in parallel when possible—this is a short investigation, and you are supposed to finish in a few seconds.
+- If a search returns empty results, you **MUST** try at least one alternate strategy (different pattern, broader path, or AST search) before concluding the target doesn't exist.
 </directives>
 
 <thoroughness>

package/src/prompts/agents/librarian.md

@@ -111,6 +111,8 @@ Before acting, determine what kind of question this is:
 - If you discover undocumented behavior or gotchas, you **MUST** populate `caveats`.
 - When local `node_modules` has the package, you **SHOULD** prefer it over cloning — it reflects the version the project actually uses.
 - You **SHOULD** use `web_search` to find the canonical repo URL and to check for known issues, but the definitive answer **MUST** come from reading source code.
+- If a search or lookup returns empty or unexpectedly few results, you **MUST** try at least 2 fallback strategies (broader query, alternate path, different source) before concluding nothing exists.
+- If the package is absent from local `node_modules` and cloning fails, you **MUST** fall back to `web_search` for official API documentation before reporting failure.
 </directives>
 
 <critical>

package/src/prompts/system/handoff-document.md

@@ -0,0 +1,46 @@
+<critical>
+Write a comprehensive handoff document for another instance of yourself.
+The handoff **MUST** be sufficient for seamless continuation without access to this conversation.
+Output ONLY the handoff document. No preamble, no commentary, no wrapper text.
+</critical>
+
+<instruction>
+Capture exact technical state, not abstractions.
+Include concrete file paths, symbol names, commands run, test results, observed failures, decisions made, and any partial work that materially affects the next step.
+</instruction>
+
+<output>
+Use exactly this structure:
+
+## Goal
+[What the user is trying to accomplish]
+
+## Constraints & Preferences
+- [Any constraints, preferences, or requirements mentioned]
+
+## Progress
+### Done
+- [x] [Completed tasks with specifics]
+
+### In Progress
+- [ ] [Current work if any]
+
+### Pending
+- [ ] [Tasks mentioned but not started]
+
+## Key Decisions
+- **[Decision]**: [Rationale]
+
+## Critical Context
+- [Code snippets, file paths, function/type names, error messages, or data essential to continue]
+- [Repository state if relevant]
+
+## Next Steps
+1. [What should happen next]
+</output>
+
+{{#if additionalFocus}}
+<instruction>
+Additional focus: {{additionalFocus}}
+</instruction>
+{{/if}}

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

@@ -29,13 +29,7 @@ Your result **MUST** match this TypeScript interface:
 {{/if}}
 
 {{SECTION_SEPERATOR "Giving Up"}}
-If you cannot complete the assignment, you **MUST** call `submit_result` exactly once with `result.error` describing what you tried and the exact blocker.
+Giving up is a last resort. If truly blocked, you **MUST** call `submit_result` exactly once with `result.error` describing what you tried and the exact blocker.
+You **MUST NOT** give up due to uncertainty, missing information obtainable via tools or repo context, or needing a design decision you can derive yourself.
 
-Giving up is a last resort.
-You **MUST NOT** give up due to uncertainty or missing information obtainable via tools or repo context.
-You **MUST NOT** give up due to requiring a design, you can derive that yourself, more than capable of that.
-
-Proceed with the best approach using the most reasonable option.
-
-You **MUST** keep going until this ticket is closed.
-This matters.
+You **MUST** keep going until this ticket is closed. This matters.

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

@@ -313,8 +313,9 @@ Today is '{{date}}', and your work begins now. Get it right.
 - You **MUST** use the most specialized tool, **NEVER** `cat` if there's tool.bash, `rg/grep`:tool.grep, `find`:tool.find, `sed`:tool.edit…
 - Every turn **MUST** materially advance the deliverable.
 - You **MUST** default to action. You **MUST NOT** ask for confirmation to continue work. If you hit an error, you **MUST** fix it. If you know the next step, you **MUST** take it. The user will intervene if needed.
-- You **MUST NOT** make speculative edits before understanding the surrounding design.
 - You **MUST** default to informed action. You **MUST NOT** ask for confirmation to continue work. If you hit an error, you **MUST** fix it. If you know the next step, you **MUST** take it. The user will intervene if needed.
+- You **MUST NOT** make speculative edits before understanding the surrounding design.
+- You **MUST NOT** stop calling tools to save round-trips when the task is incomplete. Completeness beats efficiency.
 - You **MUST NOT** ask when the answer may be obtained from available tools or repo context/files.
 - You **MUST** verify the effect. When a task involves a behavioral change, you **MUST** confirm the change is observable before yielding: run the specific test, command, or scenario that covers your change.
 </critical>

package/src/prompts/tools/bash.md

@@ -2,6 +2,8 @@ Executes bash command in shell session for terminal operations like git, bun, ca
 
 <instruction>
 - You **MUST** use `cwd` parameter to set working directory instead of `cd dir && …`
+- Prefer `env: { NAME: "…" }` for multiline, quote-heavy, or untrusted values instead of inlining them into shell syntax; reference them from the command as `$NAME`
+- Quote variable expansions like `"$NAME"` to preserve exact content and avoid shell parsing bugs
 - PTY mode is opt-in: set `pty: true` only when command expects a real terminal (for example `sudo`, `ssh` where you need input from the user); default is `false`
 - You **MUST** use `;` only when later commands should run regardless of earlier failures
 - `skill://` URIs are auto-resolved to filesystem paths before execution

package/src/session/agent-session.ts

@@ -89,6 +89,7 @@ import { getCurrentThemeName, theme } from "../modes/theme/theme";
 import { normalizeDiff, normalizeToLF, ParseError, previewPatch, stripBom } from "../patch";
 import type { PlanModeState } from "../plan-mode/state";
 import autoHandoffThresholdFocusPrompt from "../prompts/system/auto-handoff-threshold-focus.md" with { type: "text" };
+import handoffDocumentPrompt from "../prompts/system/handoff-document.md" with { type: "text" };
 import planModeActivePrompt from "../prompts/system/plan-mode-active.md" with { type: "text" };
 import planModeReferencePrompt from "../prompts/system/plan-mode-reference.md" with { type: "text" };
 import planModeToolDecisionReminderPrompt from "../prompts/system/plan-mode-tool-decision-reminder.md" with {
@@ -260,6 +261,7 @@ export interface HandoffResult {
 interface HandoffOptions {
 	autoTriggered?: boolean;
 	signal?: AbortSignal;
+	skipPostPromptRecoveryWait?: boolean;
 }
 
 /** Internal marker for hook messages queued through the agent loop */
@@ -397,7 +399,7 @@ export class AgentSession {
 	#streamingEditAbortTriggered = false;
 	#streamingEditCheckedLineCounts = new Map<string, number>();
 	#streamingEditFileCache = new Map<string, string>();
-	#promptInFlight = false;
+	#promptInFlightCount = 0;
 	#obfuscator: SecretObfuscator | undefined;
 	#pendingActionStore: PendingActionStore | undefined;
 	#checkpointState: CheckpointState | undefined = undefined;
@@ -1560,7 +1562,7 @@ export class AgentSession {
 
 	/** Whether agent is currently streaming a response */
 	get isStreaming(): boolean {
-		return this.agent.state.isStreaming || this.#promptInFlight;
+		return this.agent.state.isStreaming || this.#promptInFlightCount > 0;
 	}
 
 	/** Wait until streaming and deferred recovery work are fully settled. */
@@ -1999,7 +2001,7 @@ export class AgentSession {
 			skipPostPromptRecoveryWait?: boolean;
 		},
 	): Promise<void> {
-		this.#promptInFlight = true;
+		this.#promptInFlightCount++;
 		const generation = this.#promptGeneration;
 		try {
 			// Flush any pending bash messages before the new prompt
@@ -2109,7 +2111,7 @@ export class AgentSession {
 				await this.#waitForPostPromptRecovery();
 			}
 		} finally {
-			this.#promptInFlight = false;
+			this.#promptInFlightCount = Math.max(0, this.#promptInFlightCount - 1);
 		}
 	}
 
@@ -2508,11 +2510,10 @@ export class AgentSession {
 		this.#cancelPostPromptTasks();
 		this.agent.abort();
 		await this.agent.waitForIdle();
-		// Clear promptInFlight: waitForIdle resolves when the agent loop's finally
-		// block runs (#resolveRunningPrompt), but #promptWithMessage's finally
-		// (#promptInFlight = false) fires on a later microtask. Without this,
-		// isStreaming stays true and a subsequent prompt() throws.
-		this.#promptInFlight = false;
+		// Clear prompt-in-flight state: waitForIdle resolves when the agent loop's finally
+		// block runs, but nested prompt setup/finalizers may still be unwinding. Without this,
+		// a subsequent prompt() can incorrectly observe the session as busy after an abort.
+		this.#promptInFlightCount = 0;
 	}
 
 	/**
@@ -3194,41 +3195,9 @@ export class AgentSession {
 		}
 
 		// Build the handoff prompt
-		let handoffPrompt = `Write a comprehensive handoff document that will allow another instance of yourself to seamlessly continue this work. The document should capture everything needed to resume without access to this conversation.
-
-Use this format:
-
-## Goal
-[What the user is trying to accomplish]
-
-## Constraints & Preferences
-- [Any constraints, preferences, or requirements mentioned]
-
-## Progress
-### Done
-- [x] [Completed tasks with specifics]
-
-### In Progress
-- [ ] [Current work if any]
-
-### Pending
-- [ ] [Tasks mentioned but not started]
-
-## Key Decisions
-- **[Decision]**: [Rationale]
-
-## Critical Context
-- [Code snippets, file paths, error messages, or data essential to continue]
-- [Repository state if relevant]
-
-## Next Steps
-1. [What should happen next]
-
-Be thorough - include exact file paths, function names, error messages, and technical details. Output ONLY the handoff document, no other text.`;
-
-		if (customInstructions) {
-			handoffPrompt += `\n\nAdditional focus: ${customInstructions}`;
-		}
+		const handoffPrompt = renderPromptTemplate(handoffDocumentPrompt, {
+			additionalFocus: customInstructions,
+		});
 
 		// Create a promise that resolves when the agent completes
 		let handoffText: string | undefined;
@@ -3273,11 +3242,16 @@ Be thorough - include exact file paths, function names, error messages, and tech
 			if (handoffSignal.aborted) {
 				throw new Error("Handoff cancelled");
 			}
-			await this.prompt(handoffPrompt, {
-				expandPromptTemplates: false,
-				synthetic: true,
-				skipCompactionCheck: true,
-			});
+			await this.#promptWithMessage(
+				{
+					role: "developer",
+					content: [{ type: "text", text: handoffPrompt }],
+					attribution: "agent",
+					timestamp: Date.now(),
+				},
+				handoffPrompt,
+				{ skipCompactionCheck: true, skipPostPromptRecoveryWait: options?.skipPostPromptRecoveryWait },
+			);
 			await completionPromise;
 
 			if (handoffCancelled || handoffSignal.aborted) {
@@ -3730,11 +3704,22 @@ Be thorough - include exact file paths, function names, error messages, and tech
 	/**
 	 * Internal: Run auto-compaction with events.
 	 */
-	async #runAutoCompaction(reason: "overflow" | "threshold", willRetry: boolean): Promise<void> {
+	async #runAutoCompaction(reason: "overflow" | "threshold", willRetry: boolean, deferred = false): Promise<void> {
 		const compactionSettings = this.settings.getGroup("compaction");
-
 		if (!compactionSettings.enabled || compactionSettings.strategy === "off") return;
 		const generation = this.#promptGeneration;
+		if (!deferred && reason !== "overflow" && compactionSettings.strategy === "handoff") {
+			this.#schedulePostPromptTask(
+				async signal => {
+					await Promise.resolve();
+					if (signal.aborted) return;
+					await this.#runAutoCompaction(reason, willRetry, true);
+				},
+				{ generation },
+			);
+			return;
+		}
+
 		let action: "context-full" | "handoff" =
 			compactionSettings.strategy === "handoff" && reason !== "overflow" ? "handoff" : "context-full";
 		await this.#emitSessionEvent({ type: "auto_compaction_start", reason, action });
@@ -3750,6 +3735,7 @@ Be thorough - include exact file paths, function names, error messages, and tech
 				const handoffResult = await this.handoff(handoffFocus, {
 					autoTriggered: true,
 					signal: this.#autoCompactionAbortController.signal,
+					skipPostPromptRecoveryWait: true,
 				});
 				if (!handoffResult) {
 					const aborted = this.#autoCompactionAbortController.signal.aborted;
@@ -4028,15 +4014,25 @@ Be thorough - include exact file paths, function names, error messages, and tech
 			await this.#emitSessionEvent({ type: "auto_compaction_end", action, result, aborted: false, willRetry });
 
 			if (!willRetry && compactionSettings.autoContinue !== false) {
-				await this.#promptWithMessage(
-					{
-						role: "developer",
-						content: [{ type: "text", text: "Continue if you have next steps." }],
-						attribution: "agent",
-						timestamp: Date.now(),
+				const continuePrompt = async () => {
+					await this.#promptWithMessage(
+						{
+							role: "developer",
+							content: [{ type: "text", text: "Continue if you have next steps." }],
+							attribution: "agent",
+							timestamp: Date.now(),
+						},
+						"Continue if you have next steps.",
+						{ skipPostPromptRecoveryWait: true },
+					);
+				};
+				this.#schedulePostPromptTask(
+					async signal => {
+						await Promise.resolve();
+						if (signal.aborted) return;
+						await continuePrompt();
 					},
-					"Continue if you have next steps.",
-					{ skipPostPromptRecoveryWait: true },
+					{ generation },
 				);
 			}