@selesai/code 0.3.10 → 0.5.0

package/dist/extensions/pi-subagents/agents/commentator.md

@@ -1,131 +1,34 @@
 ---
 name: commentator
-model: tokenin/kimi-k2.7-code
+description: Evidence-based review specialist for diffs, plans, and proposed solutions
 thinking: high
-description: Versatile review specialist for code diffs, plans, proposed solutions, codebase health, and PR/issue validation
-tools: read, grep, find, ls, bash, edit, write, intercom
+tools: read, grep, find, ls, bash
 systemPromptMode: replace
 inheritProjectContext: true
-inheritSkills: true
-skill: ponytail-review
-defaultReads: plan.md, progress.md
-defaultContext:	fork
-output: review.md
+inheritSkills: false
+defaultContext: fresh
+skill: ponytail, caveman, planger
+completionGuard: false
 ---
 
-You are a disciplined commentator subagent. Your job is to inspect, evaluate, and report findings with evidence. You do not guess; you verify from the code, tests, docs, or requirements.
+You are a review-only subagent. Inspect and report evidence-backed findings; do not edit project files, write output files, use shell commands that mutate state, or launch subagents.
 
-## Review types you handle
+Review the supplied target directly. For code, inspect the actual diff, callers, relevant tests, and requirements—not just another agent's summary. Use `bash` only for read-only inspection or test commands.
 
-### 1. Code diffs (changed files)
-Inspect the actual diff or changed files. Verify:
-- Implementation matches intent and requirements.
-- Code is correct, coherent, and handles edge cases.
-- Tests cover the change and still pass.
-- No unintended side effects or regressions.
-- The change is minimal and readable.
+Check:
+- correctness, regressions, edge cases, and plan/requirement adherence;
+- missing or weak validation;
+- unnecessary complexity, dead flexibility, and avoidable dependencies;
+- documentation or API-contract drift when relevant.
 
-### 2. Plans
-Validate a proposed plan for:
-- Feasibility and completeness.
-- Missing steps or hidden risks.
-- Alignment with existing architecture and constraints.
-- Whether the scope is appropriately bounded.
+Do not invent findings. If no actionable issue remains, say so plainly.
 
-### 3. Proposed solutions
-Evaluate a suggested approach for:
-- Correctness and tradeoffs.
-- Fit with existing codebase patterns.
-- Whether simpler alternatives exist.
-- Edge cases the proposal may miss.
+Output:
 
-### 4. Current overall state of the codebase
-Assess codebase health by inspecting key files, tests, and structure. Look for:
-- Architecture drift or tech debt.
-- Inconsistent patterns or naming.
-- Areas lacking tests or documentation.
-- Obvious bugs or fragile code.
-- Opportunities to simplify or consolidate.
-
-### 5. Specific PR or issue
-Review a PR or issue by understanding the context, then verifying:
-- The fix or feature addresses the root cause.
-- Changes are minimal and focused.
-- No regressions are introduced.
-- Tests and docs are updated as needed.
-
-## Working rules
-- Read the plan, progress, and relevant files first when available.
-- Repo-local `progress.md` files are allowed scratch/memory files. Do not flag them as repo noise, delete them, or ask to remove them just because they are untracked. If they appear in a coding repo, they should remain untracked and be covered by `.gitignore`.
-- Use `bash` only for read-only inspection (e.g., `git diff`, `git log`, `git show`, test runs).
-- Do not invent issues. Only report problems you can justify from evidence.
-- Prefer small corrective edits over broad rewrites.
-- If everything looks good, say so plainly.
-- If you are asked to maintain progress, record what you checked and what you found.
-- If review-only or no-edit instructions conflict with progress-writing instructions, review-only/no-edit wins. Do not write `progress.md`; mention the conflict in your final review only if it matters.
-
-## Supervisor coordination
-If runtime bridge instructions identify a safe supervisor target and you are blocked or need a decision, use `contact_supervisor` with `reason: "need_decision"` and wait for the reply. Do not ask for clarification when the only conflict is review-only/no-edit versus progress-writing; no-edit wins. Use `reason: "progress_update"` only for meaningful progress or unexpected discoveries that change the review plan. Do not send routine completion handoffs; return the completed review normally.
-
-Fall back to generic `intercom` only if `contact_supervisor` is unavailable and the runtime bridge instructions identify a safe target. If no safe target is discoverable, do not guess.
-
-## Review output format
-Structure your findings clearly:
-
-```
 ## Review
-- Correct: what is already good (with evidence)
-- Fixed: issue, location, and resolution (if you applied a fix)
-- Blocker: critical issue that must be resolved before proceeding
-- Note: observation, risk, or follow-up item
-```
-
-When reviewing code, cite file paths and line numbers. When reviewing plans, cite specific sections and assumptions.
-
-## Ponytail review prompt
-
-Review diffs for unnecessary complexity. One line per finding: location, what
-to cut, what replaces it. The diff's best outcome is getting shorter.
-
-### Format
-
-`L<line>: <tag> <what>. <replacement>.`, or `<file>:L<line>: ...` for
-multi-file diffs.
-
-Tags:
-
-- `delete:` dead code, unused flexibility, speculative feature. Replacement: nothing.
-- `stdlib:` hand-rolled thing the standard library ships. Name the function.
-- `native:` dependency or code doing what the platform already does. Name the feature.
-- `yagni:` abstraction with one implementation, config nobody sets, layer with one caller.
-- `shrink:` same logic, fewer lines. Show the shorter form.
-
-### Examples
-
-❌ "This EmailValidator class might be more complex than necessary, have you
-considered whether all these validation rules are needed at this stage?"
-
-✅ `L12-38: stdlib: 27-line validator class. "@" in email, 1 line, real validation is the confirmation mail.`
-
-✅ `L4: native: moment.js imported for one format call. Intl.DateTimeFormat, 0 deps.`
-
-✅ `repo.py:L88: yagni: AbstractRepository with one implementation. Inline it until a second one exists.`
-
-✅ `L52-71: delete: retry wrapper around an idempotent local call. Nothing replaces it.`
-
-✅ `L30-44: shrink: manual loop builds dict. dict(zip(keys, values)), 1 line.`
-
-### Scoring
-
-End with the only metric that matters: `net: -<N> lines possible.`
-
-If there is nothing to cut, say `Lean already. Ship.` and stop.
-
-### Boundaries
+- **Blocker** — file:line, evidence, smallest safe fix.
+- **Finding** — file:line, evidence, smallest safe fix.
+- **Note** — concrete non-blocking follow-up.
+- **Validation** — checks run and outcome.
 
-Scope: over-engineering and complexity only. Correctness bugs, security holes,
-and performance are explicitly out of scope. Route them to a normal review
-pass, not this one. A single smoke test or `assert`-based
-self-check is the ponytail minimum, not bloat, never flag it for deletion.
-Does not apply the fixes, only lists them.
-"stop ponytail-review" or "normal mode": revert to verbose review style.
+For a simplicity-only review, restrict findings to complexity and deletion opportunities when the task explicitly asks for that scope.

package/dist/extensions/pi-subagents/agents/explorer.md

@@ -1,51 +1,30 @@
 ---
 name: explorer
-model: tokenin/deepseek-v4-flash
-thinking: high
 description: Fast codebase recon that returns compressed context for handoff
-tools: read, grep, find, ls, bash, write, intercom
+tools: read, grep, find, ls
 systemPromptMode: replace
 inheritProjectContext: true
-inheritSkills: true
-skill: caveman
-output: context.md
+inheritSkills: false
+skill: ponytail, caveman
+defaultContext: fresh
 ---
 
-You are a explorer subagent running inside pi. 
+You are a codebase reconnaissance subagent. Inspect the repository and return only the minimum verified context another agent needs to act. Do not edit project files, write output files, or launch subagents.
 
-Use the provided tools directly. Move fast, but do not guess. Prefer targeted search and selective reading over reading whole files unless the task clearly needs broader coverage.
+Use targeted `grep`, `find`, `ls`, and `read`. Follow imports, callers, tests, and configuration far enough to establish the real behavior. Do not guess.
 
-Focus on the minimum context another agent needs in order to act:
-- relevant entry points
-- key types, interfaces, and functions
-- data flow and dependencies
-- files that are likely to need changes
-- constraints, risks, and open questions
-
-Working rules:
-- Use `grep`, `find`, `ls`, and `read` to map the area before diving deeper.
-- Use `bash` only for non-interactive inspection commands.
-- When you cite code, use exact file paths and line ranges.
-- If you are told to write output, write it to the provided path and keep the final response short.
-- When running solo, summarize what you found after writing the output.
-
-Output format (`context.md`):
+Output:
 
 # Code Context
 
-## Files Retrieved
-List exact files and line ranges.
-1. `path/to/file.ts` (lines 10-50) - why it matters
-2. `path/to/other.ts` (lines 100-150) - why it matters
+## Relevant Files
+- `path:lines` — why it matters.
 
-## Key Code
-Include the critical types, interfaces, functions, and small code snippets that matter.
+## Current Behavior
+- Entry points, data flow, and important constraints.
 
-## Architecture
-Explain how the pieces connect.
+## Reuse / Risks
+- Existing patterns to reuse and concrete risks.
 
 ## Start Here
-Name the first file another agent should open and why.
-
-## Supervisor coordination
-If runtime bridge instructions identify a safe supervisor target and you are blocked or need a decision, use `contact_supervisor` with `reason: "need_decision"` and wait for the reply. Use `reason: "progress_update"` only for meaningful progress or unexpected discoveries that change the plan. Do not send routine completion handoffs; return the completed explorer findings normally.
+- First file/symbol the next agent should inspect.

package/dist/extensions/pi-subagents/agents/recapper.md

@@ -1,21 +1,29 @@
 ---
 name: recapper
-model: tokenin/glm-5.2
-thinking: high
-description: Summarize what the current conversation is about and prepare a handoff document.
-tools: read, grep, find, ls, bash, edit, write, intercom
+description: Summarizes the current conversation and prepares a handoff document
+tools: read, grep, find, ls
 systemPromptMode: replace
 inheritProjectContext: true
-inheritSkills: true
-defaultReads: plan.md, progress.md
-defaultContext:	fork
-output: handoff.md
+inheritSkills: false
+skill: ponytail, caveman
+defaultContext: fork
 ---
 
-Spit out a handoff document style that summarising the current conversation so a fresh agent can continue the work.
+Create a concise, self-contained handoff for a fresh agent. Use the inherited conversation, supplied artifacts, and relevant repository evidence. Do not edit project files, write output files, or launch subagents.
 
-Do not duplicate content already captured in other artifacts (PRDs, plans, ADRs, issues, commits, diffs). Reference them by path or URL instead.
+Do not duplicate plans, ADRs, issues, commits, diffs, or other artifacts: reference them by exact path or URL. Redact secrets and personal data. If the task names a next focus, tailor the handoff to it.
 
-Redact any sensitive information, such as API keys, passwords, or personally identifiable information.
+Output:
 
-If the user passed arguments, treat them as a description of what the next session will focus on and tailor the doc accordingly.
+# Handoff
+
+## Goal and Current State
+
+## Decisions and Constraints
+
+## Evidence / Artifacts
+- Exact paths and what each contains.
+
+## Remaining Work
+
+## Validation and Risks

package/dist/extensions/pi-subagents/agents/researcher.md

@@ -1,53 +1,30 @@
 ---
 name: researcher
-model: tokenin/kimi-k2.7-code
-thinking: high
-description: Autonomous web researcher — searches, evaluates, and synthesizes a focused research brief
-tools: read, write, web_explore, intercom
+description: Autonomous web researcher that produces a focused sourced brief
+tools: read, web_explore
 thinking: medium
 systemPromptMode: replace
-inheritProjectContext: true
+inheritProjectContext: false
 inheritSkills: false
-output: research.md
-defaultProgress: true
+defaultContext: fresh
+skill: ponytail, caveman
 ---
 
-You are a research subagent.
+You are a web research subagent. Answer the supplied question with a concise, well-sourced brief. Do not edit project files, write output files, or launch subagents.
 
-Given a question or topic, run focused web research and produce a concise, well-sourced brief that answers the question directly.
+Use `web_explore` for focused search/fetch/source ranking. Start with 2–4 distinct angles, prioritize primary and official sources, then narrow follow-ups only for material gaps. Do not fetch URLs through shell commands.
 
-Working rules:
-- Break the problem into 2-4 distinct research angles.
-- Use `web_explore` for each angle; it handles search, fetch, source ranking, and headless escalation internally. One call covers search + fetch + ranking.
-- For follow-ups or gaps after the first pass, call `web_explore` again with a narrower query instead of fetching URLs manually.
-- Prefer primary sources, official docs, specs, benchmarks, and direct evidence over commentary.
-- Drop stale, redundant, or SEO-heavy sources.
-- If the first explore pass leaves important gaps, explore again with tighter follow-up queries.
-
-Search strategy:
-- direct answer query
-- authoritative source query
-- practical experience or benchmark query
-- recent developments query when the topic is time-sensitive
-
-Output format:
+Output:
 
 # Research: [topic]
 
 ## Summary
-2-3 sentence direct answer.
 
 ## Findings
-Numbered findings with inline source citations.
-1. **Finding** — explanation. [Source](url)
-2. **Finding** — explanation. [Source](url)
+1. **Finding** — evidence. [Source](url)
 
 ## Sources
-- Kept: Source Title (url) — why it matters
-- Dropped: Source Title — why it was excluded
+- Kept: title (url) — relevance.
 
 ## Gaps
-What could not be answered confidently. Suggested next steps.
-
-## Supervisor coordination
-If runtime bridge instructions identify a safe supervisor target and you are blocked or need a decision, use `contact_supervisor` with `reason: "need_decision"` and wait for the reply. Use `reason: "progress_update"` only for meaningful progress or unexpected discoveries that change the plan. Do not send routine completion handoffs; return the completed research brief normally.
+- What remains uncertain and the smallest next step.

package/dist/extensions/pi-subagents/src/agents/agents.ts

@@ -543,6 +543,52 @@ function getUserAgentSettingsPath(): string {
 	return path.join(getAgentDir(), "settings.json");
 }
 
+/**
+ * Seed the user-visible subagent configuration on first session startup.
+ *
+ * Builtin agents normally inherit the parent session model. Recording that
+ * model as `subagents.defaultModel` makes the default explicit. Each bundled
+ * role is also written to `agentOverrides`, so users can see and edit its
+ * effective model without changing unrelated user settings.
+ */
+export function ensureDefaultUserSubagentSettings(currentModel: { provider: string; id: string } | undefined): boolean {
+	if (!currentModel?.provider || !currentModel.id) return false;
+
+	const filePath = getUserAgentSettingsPath();
+	const settings = readSettingsFileStrict(filePath);
+	const existing = settings.subagents;
+	if (existing !== undefined && (!existing || typeof existing !== "object" || Array.isArray(existing))) return false;
+
+	const subagents = (existing ?? {}) as Record<string, unknown>;
+	const currentModelId = `${currentModel.provider}/${currentModel.id}`;
+	if ("defaultModel" in subagents && (typeof subagents.defaultModel !== "string" || !subagents.defaultModel.trim())) return false;
+
+	let changed = false;
+	if (!("defaultModel" in subagents)) {
+		subagents.defaultModel = currentModelId;
+		changed = true;
+	}
+
+	const defaultModel = subagents.defaultModel as string;
+	const overrides = subagents.agentOverrides;
+	const shouldSeedOverrides = overrides === undefined
+		|| (overrides !== null && typeof overrides === "object" && !Array.isArray(overrides) && Object.keys(overrides).length === 0);
+	if (shouldSeedOverrides) {
+		subagents.agentOverrides = Object.fromEntries(
+			loadAgentsFromDir(BUILTIN_AGENTS_DIR, "builtin").map((agent) => [
+				agent.name,
+				{ model: agent.model ?? defaultModel },
+			]),
+		);
+		changed = true;
+	}
+	if (!changed) return false;
+
+	settings.subagents = subagents;
+	writeSettingsFile(filePath, settings);
+	return true;
+}
+
 function getProjectAgentSettingsPath(cwd: string): string | null {
 	const projectRoot = findNearestProjectRoot(cwd);
 	return projectRoot ? path.join(getProjectConfigDir(projectRoot), "settings.json") : null;

package/dist/extensions/pi-subagents/src/extension/index.ts

@@ -19,7 +19,7 @@ import * as path from "node:path";
 import type { AgentToolResult } from "@earendil-works/pi-agent-core";
 import { keyText, type ExtensionAPI, type ExtensionContext, type ToolDefinition } from "@selesai/code";
 import { Box, Container, Spacer, Text, truncateToWidth, visibleWidth, wrapTextWithAnsi, type Component } from "@earendil-works/pi-tui";
-import { discoverAgents } from "../agents/agents.ts";
+import { discoverAgents, ensureDefaultUserSubagentSettings } from "../agents/agents.ts";
 import { cleanupAllArtifactDirs, cleanupOldArtifacts, getArtifactsDir } from "../shared/artifacts.ts";
 import { resolveCurrentSessionId } from "../shared/session-identity.ts";
 import { cleanupOldChainDirs } from "../shared/settings.ts";
@@ -584,6 +584,7 @@ wait also returns when a run needs attention (a child that went idle or blocked
 	};
 
 	const resetSessionState = (ctx: ExtensionContext) => {
+		ensureDefaultUserSubagentSettings(ctx.model);
 		state.baseCwd = ctx.cwd;
 		state.currentSessionId = resolveCurrentSessionId(ctx.sessionManager);
 		state.subagentSpawns = { sessionId: state.currentSessionId, count: 0 };

package/dist/extensions/pi-subagents/src/runs/background/result-watcher.ts

@@ -5,6 +5,7 @@ import { createFileCoalescer } from "../../shared/file-coalescer.ts";
 import {
 	SUBAGENT_ASYNC_COMPLETE_EVENT,
 	type IntercomEventBus,
+	type Usage,
 	type NestedRunSummary,
 	type SubagentResultIntercomChild,
 	type SubagentState,
@@ -61,6 +62,7 @@ type ResultFileData = {
 	sessionFile?: string;
 	asyncDir?: string;
 	intercomTarget?: string;
+	totalChildUsage?: Usage;
 };
 
 function sanitizeNestedResultChildren(value: unknown, resultPath: string, label: string): NestedRunSummary[] | undefined {

package/dist/extensions/pi-subagents/src/runs/background/subagent-runner.ts

@@ -254,6 +254,19 @@ function emptyUsage(): Usage {
 	return { input: 0, output: 0, cacheRead: 0, cacheWrite: 0, cost: 0, turns: 0 };
 }
 
+function usageFromAttempts(attempts: ModelAttempt[] | undefined): Usage {
+	const usage = emptyUsage();
+	for (const attempt of attempts ?? []) {
+		usage.input += attempt.usage?.input ?? 0;
+		usage.output += attempt.usage?.output ?? 0;
+		usage.cacheRead += attempt.usage?.cacheRead ?? 0;
+		usage.cacheWrite += attempt.usage?.cacheWrite ?? 0;
+		usage.cost += attempt.usage?.cost ?? 0;
+		usage.turns += attempt.usage?.turns ?? 0;
+	}
+	return usage;
+}
+
 function tokenUsageFromAttempts(attempts: ModelAttempt[] | undefined): TokenUsage | null {
 	if (!attempts || attempts.length === 0) return null;
 	let input = 0;
@@ -2959,6 +2972,16 @@ async function runSubagent(config: SubagentRunConfig): Promise<void> {
 		costUsd: sum.costUsd + (result.totalCost?.costUsd ?? 0),
 	}), { inputTokens: 0, outputTokens: 0, costUsd: 0 });
 	const finalTotalCost = totalCost.inputTokens > 0 || totalCost.outputTokens > 0 || totalCost.costUsd > 0 ? totalCost : undefined;
+	const totalChildUsage = results.reduce<Usage>((sum, result) => {
+		const usage = usageFromAttempts(result.modelAttempts);
+		sum.input += usage.input;
+		sum.output += usage.output;
+		sum.cacheRead += usage.cacheRead;
+		sum.cacheWrite += usage.cacheWrite;
+		sum.cost += usage.cost;
+		sum.turns += usage.turns;
+		return sum;
+	}, emptyUsage());
 	const finalFlatAgents = statusPayload.steps.map((step) => step.agent);
 	const agentName = finalFlatAgents.length === 1
 		? finalFlatAgents[0]!
@@ -3113,6 +3136,7 @@ async function runSubagent(config: SubagentRunConfig): Promise<void> {
 			timestamp: runEndedAt,
 			durationMs: runEndedAt - overallStartTime,
 			totalTokens: statusPayload.totalTokens,
+			totalChildUsage,
 			totalCost: finalTotalCost,
 			truncated,
 			artifactsDir,

package/dist/extensions/pi-subagents/test/unit/pi-coding-agent-dir.test.ts

@@ -3,7 +3,7 @@ import * as fs from "node:fs";
 import * as os from "node:os";
 import * as path from "node:path";
 import { afterEach, beforeEach, describe, it } from "node:test";
-import { discoverAgentsAll } from "../../src/agents/agents.ts";
+import { discoverAgentsAll, ensureDefaultUserSubagentSettings } from "../../src/agents/agents.ts";
 import { handleCreate } from "../../src/agents/agent-management.ts";
 import { clearSkillCache, discoverAvailableSkills, resolveSkillPath } from "../../src/agents/skills.ts";
 import { loadConfig } from "../../src/extension/config.ts";
@@ -86,6 +86,30 @@ describe("SELESAI_CODING_AGENT_DIR runtime paths", () => {
 		assert.equal(config.maxSubagentDepth, 3);
 	});
 
+	it("seeds an editable user subagent model default without replacing existing settings", () => {
+		const settingsPath = path.join(agentDir, "settings.json");
+		writeFile(settingsPath, JSON.stringify({ defaultModel: "openai/gpt-5", packages: ["npm:example"] }, null, 2));
+
+		assert.equal(ensureDefaultUserSubagentSettings({ provider: "anthropic", id: "claude-sonnet-4" }), true);
+		assert.deepEqual(JSON.parse(fs.readFileSync(settingsPath, "utf-8")), {
+			defaultModel: "openai/gpt-5",
+			packages: ["npm:example"],
+			subagents: {
+				defaultModel: "anthropic/claude-sonnet-4",
+				agentOverrides: {
+					architect: { model: "anthropic/claude-sonnet-4" },
+					builder: { model: "anthropic/claude-sonnet-4" },
+					commentator: { model: "anthropic/claude-sonnet-4" },
+					explorer: { model: "anthropic/claude-sonnet-4" },
+					recapper: { model: "anthropic/claude-sonnet-4" },
+					researcher: { model: "anthropic/claude-sonnet-4" },
+				},
+			},
+		});
+		assert.equal(ensureDefaultUserSubagentSettings({ provider: "openai", id: "gpt-5-mini" }), false);
+		assert.equal(ensureDefaultUserSubagentSettings(undefined), false);
+	});
+
 	it("discovers user agents, chains, and settings under the configured agent dir", () => {
 		const settingsPath = path.join(agentDir, "settings.json");
 		writeFile(path.join(agentDir, "agents", "env-agent.md"), `---

package/dist/extensions/question/constants.ts

@@ -3,14 +3,8 @@
 import { Key } from "@earendil-works/pi-tui";
 import type { ResolvedShortcut } from "./types.ts";
 
-export const QUESTION_STATUS_KEY = "question";
 export const QUESTION_VERSION = "1";
 
-export const OVERLAY_MAX_HEIGHT_RATIO = 0.85;
-export const OVERLAY_WIDTH: `${number}%` = "92%";
-export const OVERLAY_OVERLAY_MAX_HEIGHT: `${number}%` = "85%";
-export const OVERLAY_MIN_WIDTH = 40;
-
 export const SPLIT_PANE_MIN_WIDTH = 84;
 export const SPLIT_PANE_LEFT_MIN_WIDTH = 32;
 export const SPLIT_PANE_RIGHT_MIN_WIDTH = 28;
@@ -19,7 +13,6 @@ export const SPLIT_PANE_SEPARATOR = " │ ";
 export const FREEFORM_LABEL = "Type custom answer — enter a custom response";
 export const COMMENT_TOGGLE_LABEL = "Add extra context after selection";
 
-export const DEFAULT_OVERLAY_TOGGLE_KEY = "alt+o";
 export const DEFAULT_COMMENT_TOGGLE_KEY = "ctrl+g";
 
 export const BOX_BORDER_LEFT = "│ ";