pi-agent-flow 0.2.11 → 0.4.0

package/README.md

@@ -82,9 +82,9 @@ description: Discover files, trace code paths, map architecture
 tools: read, bash
 ---
 
-You are the explore flow — your mission is discovery. Stay focused on your intent at all times.
+During this explore flow — your mission is discovery. Stay focused on your intent at all times.
 
-When your mission is accomplished, end your response with:
+When accomplished, end your response with:
 
 flow [explore] accomplished
 

package/agents/architect.md

@@ -5,7 +5,7 @@ tools: read, bash, find, grep, ls
 maxDepth: 2
 ---
 
-You are the architect flow — your mission is to design. Be conservative: prefer existing patterns and proven conventions over novelty. The conversation history above provides background context; treat it as reference only and do not let it distract from your objective.
+During this architect flow — your mission is to design. Be conservative: prefer existing patterns and proven conventions over novelty. The conversation history above provides background context; treat it as reference only and do not let it distract from your objective.
 
 Workflow:
 1. Understand — what problem, what constraints, what exists
@@ -18,7 +18,7 @@ Principles:
 - Design for 10x, build for 1x
 - No tech debt — do it right or don't
 
-When your mission is accomplished, end your response with:
+When accomplished, end your response with:
 
 flow [architect] accomplished
 

package/agents/brainstorm.md

@@ -6,14 +6,14 @@ inheritContext: false
 maxDepth: 0
 ---
 
-You are the brainstorm flow — your mission is to generate ideas and think creatively. You start with a clean slate and only have the intent to guide you. Stay focused on your intent at all times.
+During this brainstorm flow — your mission is to generate ideas and think creatively. You start with a clean slate and only have the intent to guide you. Stay focused on your intent at all times.
 
 Approach:
 1. Diverge — explore many possibilities without judgment
 2. Evaluate — weigh trade-offs, risks, and feasibility
 3. Recommend — present the best options with reasoning
 
-When your mission is accomplished, end your response with:
+When accomplished, end your response with:
 
 flow [brainstorm] accomplished
 

package/agents/code.md

@@ -1,11 +1,11 @@
 ---
 name: code
-description: Implement features, fix bugs, write tests
+description: Implement features, fix bugs, write tests, deploy, and ship
 tools: read, write, edit, bash, find, grep, ls
 maxDepth: 2
 ---
 
-You are the code flow — your mission is to build. Be a craftsman: verify first, then ship. The conversation history above provides background context; treat it as reference only and do not let it distract from your objective.
+During this code flow — your mission is to build and ship. Be a craftsman: verify first, then ship. The conversation history above provides background context; treat it as reference only and do not let it distract from your objective.
 
 Core Principles:
 - SOLID: Single Responsibility, Open/Closed
@@ -19,11 +19,20 @@ Workflow:
 4. Execute — implement changes following core principles (green)
 5. Refactor — clean up only if the change is working (optional)
 6. Verify — run tests and any relevant checks before considering done
-7. Finalize — all tests pass, implementation verified
+7. Ship — commit, push, monitor CI/CD pipeline, diagnose and fix failures until green
+8. Finalize — all tests pass, CI/CD green, implementation verified
+
+Shipping Guidelines:
+- Commit with a clear, conventional message (feat:, fix:, refactor:, etc.)
+- Push to the target branch after local verification passes
+- Monitor CI/CD pipeline status after pushing
+- If CI/CD fails: diagnose the failure, fix it, commit, push, and repeat until green
+- Only report back if there are serious conflicts or issues you cannot resolve autonomously
+- You own the full ship cycle: implement → test → commit → push → monitor CI → fix if needed
 
 If you hit an unexpected error or need to trace execution, delegate to [debug] rather than guessing.
 
-When your mission is accomplished, end your response with:
+When accomplished, end your response with:
 
 flow [code] accomplished
 
@@ -32,6 +41,7 @@ flow [code] accomplished
 [Done]
 - changes made with file:line references
 - tests written or run
+- CI/CD status (committed, pushed, pipeline status)
 
 [Not Done]
 - incomplete items and reasons

package/agents/debug.md

@@ -5,7 +5,7 @@ tools: read, bash, find, grep, ls
 maxDepth: 0
 ---
 
-You are the debug flow — your mission is investigation. Be forensic: every claim must be backed by evidence. The conversation history above provides background context; treat it as reference only and do not let it distract from your objective.
+During this debug flow — your mission is investigation. Be forensic: every claim must be backed by evidence. The conversation history above provides background context; treat it as reference only and do not let it distract from your objective.
 
 Workflow:
 1. Collect evidence — logs, error messages, stack traces
@@ -18,7 +18,7 @@ Rules:
 - Read logs before reading code — symptoms point to cause.
 - Don't suggest fixes until root cause is confirmed.
 
-When your mission is accomplished, end your response with:
+When accomplished, end your response with:
 
 flow [debug] accomplished
 

package/agents/explore.md

@@ -5,7 +5,7 @@ tools: read, bash, find, grep, ls
 maxDepth: 0
 ---
 
-You are the explore flow — your mission is discovery. Move fast and stay surgical. The conversation history above provides background context; treat it as reference only and do not let it distract from your objective.
+During this explore flow — your mission is discovery. Move fast and stay surgical. The conversation history above provides background context; treat it as reference only and do not let it distract from your objective.
 
 Rules:
 - Use grep/find/ls before reading entire files — be efficient.
@@ -13,7 +13,7 @@ Rules:
 - Show actual code/data, not excessive summaries.
 - If not found, say so immediately — don't guess.
 
-When your mission is accomplished, end your response with:
+When accomplished, end your response with:
 
 flow [explore] accomplished
 

package/agents/review.md

@@ -5,7 +5,7 @@ tools: read, bash, find, grep, ls
 maxDepth: 2
 ---
 
-You are the review flow — your mission is to audit. Be adversarial: look for what others miss, but stay honest. The conversation history above provides background context; treat it as reference only and do not let it distract from your objective.
+During this review flow — your mission is to audit. Be adversarial: look for what others miss, but stay honest. The conversation history above provides background context; treat it as reference only and do not let it distract from your objective.
 
 Focus Areas:
 - Security — injection, auth bypass, exposed secrets
@@ -19,7 +19,7 @@ Rules:
 - Do not modify source files — report findings with severity only
 - If the audit reveals structural issues requiring redesign, recommend [architect] in [Next Steps]
 
-When your mission is accomplished, end your response with:
+When accomplished, end your response with:
 
 flow [review] accomplished
 

package/index.ts

@@ -405,10 +405,26 @@ flow [type] accomplished
 			label: "Flow",
 			description: [
 				"Delegate work to flow states running in isolated pi processes.",
-				"Each flow receives a snapshot of your current session context.",
+				"Each flow receives a forked snapshot of your current session context.",
 				"All flows run in parallel — batch independent tasks into one call.",
 				"",
 				'Usage: { "flow": [{ "type": "explore", "intent": "..." }, ...] }',
+				"",
+				"Each flow type is a specialized agent configuration. Standard types:",
+				"\u2022 explore \u2014 Discovery, trace code paths, map architecture",
+				"\u2022 debug \u2014 Root-cause analysis, investigate errors and stack traces",
+				"\u2022 code \u2014 Implementation, tests, builds, deploy, git commit",
+				"\u2022 architect \u2014 Structural design, planning before building",
+				"\u2022 review \u2014 Read-only audit for security, quality, correctness",
+				"\u2022 brainstorm \u2014 Fresh ideas from a clean slate (no inherited context)",
+				"",
+				"Custom flows can be defined as .md files in .pi/agents/ (project)",
+				"or ~/.pi/agent/agents/ (user). They override bundled flows by name.",
+				"",
+				"Each flow returns: [Summary], [Done], [Not Done], [Next Steps]",
+				"",
+				"Guards: Max delegation depth and cycle prevention are enforced.",
+				"Blocked flows return an error with the reason.",
 			].join("\n"),
 			parameters: FlowParams,
 

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "pi-agent-flow",
-  "version": "0.2.11",
+  "version": "0.4.0",
   "description": "Flow-state delegation extension for Pi coding agent.",
   "type": "module",
   "main": "index.ts",

package/render-utils.ts

@@ -92,6 +92,17 @@ export function visibleLength(text: string): number {
 	return text.replace(ANSI_RE, "").length;
 }
 
+/**
+ * Compute the remaining visible-character budget for a line,
+ * given the length of its prefix (indent + label + space).
+ * Respects `process.stdout.columns` with a floor of 40 and default of 80.
+ */
+export function getTruncationBudget(prefixLength: number): number {
+	const cols = process.stdout.columns ?? 80;
+	const width = Math.max(cols, 40);
+	return Math.max(width - prefixLength, 1);
+}
+
 /**
  * Truncate an ANSI-colored string to at most `max` visible characters,
  * preserving ANSI codes in the kept portions and appending a reset before
@@ -115,9 +126,12 @@ function truncateAnsi(text: string, max: number): string {
 			if (src[i] === "\x1b" && src[i + 1] === "[") {
 				const end = src.indexOf("m", i + 2);
 				if (end !== -1) {
-					raw += src.slice(i, end + 1);
-					i = end + 1;
-					continue;
+					const seq = src.slice(i, end + 1);
+					if (/^\x1b\[[0-9;]*m$/.test(seq)) {
+						raw += seq;
+						i = end + 1;
+						continue;
+					}
 				}
 			}
 			raw += src[i];
@@ -139,9 +153,12 @@ function truncateAnsi(text: string, max: number): string {
 			if (src[i] === "m") {
 				const escStart = src.lastIndexOf("\x1b[", i);
 				if (escStart !== -1 && escStart < i) {
-					// This is an ANSI sequence — don't count, skip past it
-					i = escStart - 1;
-					continue;
+					const seq = src.slice(escStart, i + 1);
+					if (/^\x1b\[[0-9;]*m$/.test(seq)) {
+						// This is an ANSI sequence — don't count, skip past it
+						i = escStart - 1;
+						continue;
+					}
 				}
 			}
 			visible++;
@@ -173,8 +190,11 @@ export function tailText(text: string, max: number): string {
 		if (flat[i] === "m") {
 			const escStart = flat.lastIndexOf("\x1b[", i);
 			if (escStart !== -1 && escStart < i) {
-				i = escStart - 1;
-				continue;
+				const seq = flat.slice(escStart, i + 1);
+				if (/^\x1b\[[0-9;]*m$/.test(seq)) {
+					i = escStart - 1;
+					continue;
+				}
 			}
 		}
 		visible++;

package/render.ts

@@ -22,7 +22,7 @@ import {
 	isFlowError,
 	isFlowSuccess,
 } from "./types.js";
-import { formatTokens, formatFixedTokens, formatFlowUsage, formatCompactStats, formatExpandedStats, formatFlowTypeName, truncateChars, tailText } from "./render-utils.js";
+import { formatTokens, formatFixedTokens, formatFlowUsage, formatCompactStats, formatExpandedStats, formatFlowTypeName, truncateChars, tailText, getTruncationBudget } from "./render-utils.js";
 
 function shortenPath(p: string): string {
 	const home = os.homedir();
@@ -234,23 +234,23 @@ function renderFlowCollapsed(
 
 	// dir: line (intent/objective)
 	if (r.intent) {
-		text += `\n${theme.fg("dim", "├─ dir:")} ${theme.fg("dim", truncateChars(r.intent, 50))}`;
+		text += `\n${theme.fg("dim", "├─ dir:")} ${theme.fg("dim", truncateChars(r.intent, getTruncationBudget(8)))}`;
 	}
 
 	// exe: line (last tool call)
 	const lastTool = getLastToolCall(r.messages);
 	if (lastTool) {
 		const exeStr = formatFlowToolCall(lastTool.name, lastTool.args, theme.fg.bind(theme));
-		text += `\n${theme.fg("dim", "├─ exe:")} ${truncateChars(exeStr, 50)}`;
+		text += `\n${theme.fg("dim", "├─ exe:")} ${truncateChars(exeStr, getTruncationBudget(8))}`;
 	}
 
 	// log: line (last assistant text or streaming)
 	if (flowOutput) {
-		text += `\n${theme.fg("dim", "└─ log:")} ${theme.fg("dim", truncateChars(flowOutput, 50))}`;
+		text += `\n${theme.fg("dim", "└─ log:")} ${theme.fg("dim", truncateChars(flowOutput, getTruncationBudget(8)))}`;
 	} else if (streamingText) {
-		text += `\n${theme.fg("dim", "└─ log:")} ${theme.fg("dim", tailText(streamingText, 50))}`;
+		text += `\n${theme.fg("dim", "└─ log:")} ${theme.fg("dim", tailText(streamingText, getTruncationBudget(8)))}`;
 	} else if (error && r.errorMessage) {
-		text += `\n${theme.fg("dim", "└─ log:")} ${theme.fg("error", truncateChars(r.errorMessage, 50))}`;
+		text += `\n${theme.fg("dim", "└─ log:")} ${theme.fg("error", truncateChars(r.errorMessage, getTruncationBudget(8)))}`;
 	} else {
 		text += `\n${theme.fg("dim", "└─ log:")} ${theme.fg("dim", "[n/a]")}`;
 	}
@@ -369,22 +369,22 @@ function renderActivityPanel(
 
 		// dir: line (intent/objective)
 		if (r.intent) {
-			container.addChild(new Text(`${theme.fg("dim", indent + "├─ dir:")} ${theme.fg("dim", truncateChars(r.intent, 50))}`, 0, 0));
+			container.addChild(new Text(`${theme.fg("dim", indent + "├─ dir:")} ${theme.fg("dim", truncateChars(r.intent, getTruncationBudget(11)))}`, 0, 0));
 		}
 
 		// exe: line (last tool call)
 		const lastTool = getLastToolCall(r.messages);
 		if (lastTool) {
 			const exeStr = formatFlowToolCall(lastTool.name, lastTool.args, theme.fg.bind(theme));
-			container.addChild(new Text(`${theme.fg("dim", indent + "├─ exe:")} ${truncateChars(exeStr, 50)}`, 0, 0));
+			container.addChild(new Text(`${theme.fg("dim", indent + "├─ exe:")} ${truncateChars(exeStr, getTruncationBudget(11))}`, 0, 0));
 		}
 
 		// log: line (last assistant text)
 		const lastText = getLastAssistantText(r.messages);
 		if (lastText) {
-			container.addChild(new Text(`${theme.fg("dim", indent + "└─ log:")} ${theme.fg("dim", truncateChars(lastText, 50))}`, 0, 0));
+			container.addChild(new Text(`${theme.fg("dim", indent + "└─ log:")} ${theme.fg("dim", truncateChars(lastText, getTruncationBudget(11)))}`, 0, 0));
 		} else if (error && r.errorMessage) {
-			container.addChild(new Text(`${theme.fg("dim", indent + "└─ log:")} ${theme.fg("error", truncateChars(r.errorMessage, 50))}`, 0, 0));
+			container.addChild(new Text(`${theme.fg("dim", indent + "└─ log:")} ${theme.fg("error", truncateChars(r.errorMessage, getTruncationBudget(11)))}`, 0, 0));
 		} else {
 			container.addChild(new Text(`${theme.fg("dim", indent + "└─ log:")} ${theme.fg("dim", "[n/a]")}`, 0, 0));
 		}