jeo-code 0.4.7 → 0.4.9

package/README.ja.md

@@ -150,11 +150,11 @@ CI は `.github/workflows/npm-publish.yml` で公開します — GitHub リリ
 ## 変更履歴 (Changelog)
 
 <!-- CHANGELOG:START (auto-generated from CHANGELOG.md — run `bun run changelog:sync`) -->
+- **[0.4.9]** (2026-06-14) — Live-frame width-clamp (content-sized height) replaces the constant-height approach, typed text shows during a running turn, and a docs/AGENTS refresh.
+- **[0.4.8]** (2026-06-14) — Live-frame stability: constant-height live turn, renderer self-heal off-by-one fix, and frame-safe child-stdout sanitizing — no more duplicate model bar or torn escapes.
 - **[0.4.7]** (2026-06-14) — Detached subagents + `subagent` control tool, live shaded in-flight output, registry-driven providers, fuller `read` budget, styled italics in the final report, and `gjc` retired.
 - **[0.4.6]** (2026-06-14) — Width-correct forge cards for CJK/emoji, red borders on failed tool cards, aligned `ooo ralph` monitor HUD, and a per-theme user-card palette.
 - **[0.4.5]** (2026-06-14) — First-class filesystem make/remove tools.
-- **[0.4.4]** (2026-06-13) — Live subagent status mirroring, always-useful Ctrl+O activity tail, read lineRange crash guard.
-- **[0.4.3]** (2026-06-13) — Readability pass for autopilot, subagent activity, and worked-history review.
 
 See [CHANGELOG.md](CHANGELOG.md) for the full history.
 <!-- CHANGELOG:END -->

package/README.ko.md

@@ -150,11 +150,11 @@ CI는 `.github/workflows/npm-publish.yml`로 배포합니다 — GitHub 릴리
 ## 변경 이력 (Changelog)
 
 <!-- CHANGELOG:START (auto-generated from CHANGELOG.md — run `bun run changelog:sync`) -->
+- **[0.4.9]** (2026-06-14) — Live-frame width-clamp (content-sized height) replaces the constant-height approach, typed text shows during a running turn, and a docs/AGENTS refresh.
+- **[0.4.8]** (2026-06-14) — Live-frame stability: constant-height live turn, renderer self-heal off-by-one fix, and frame-safe child-stdout sanitizing — no more duplicate model bar or torn escapes.
 - **[0.4.7]** (2026-06-14) — Detached subagents + `subagent` control tool, live shaded in-flight output, registry-driven providers, fuller `read` budget, styled italics in the final report, and `gjc` retired.
 - **[0.4.6]** (2026-06-14) — Width-correct forge cards for CJK/emoji, red borders on failed tool cards, aligned `ooo ralph` monitor HUD, and a per-theme user-card palette.
 - **[0.4.5]** (2026-06-14) — First-class filesystem make/remove tools.
-- **[0.4.4]** (2026-06-13) — Live subagent status mirroring, always-useful Ctrl+O activity tail, read lineRange crash guard.
-- **[0.4.3]** (2026-06-13) — Readability pass for autopilot, subagent activity, and worked-history review.
 
 See [CHANGELOG.md](CHANGELOG.md) for the full history.
 <!-- CHANGELOG:END -->

package/README.md

@@ -150,11 +150,11 @@ Required npm token permissions (repository secret `NPM_TOKEN`):
 ## Changelog
 
 <!-- CHANGELOG:START (auto-generated from CHANGELOG.md — run `bun run changelog:sync`) -->
+- **[0.4.9]** (2026-06-14) — Live-frame width-clamp (content-sized height) replaces the constant-height approach, typed text shows during a running turn, and a docs/AGENTS refresh.
+- **[0.4.8]** (2026-06-14) — Live-frame stability: constant-height live turn, renderer self-heal off-by-one fix, and frame-safe child-stdout sanitizing — no more duplicate model bar or torn escapes.
 - **[0.4.7]** (2026-06-14) — Detached subagents + `subagent` control tool, live shaded in-flight output, registry-driven providers, fuller `read` budget, styled italics in the final report, and `gjc` retired.
 - **[0.4.6]** (2026-06-14) — Width-correct forge cards for CJK/emoji, red borders on failed tool cards, aligned `ooo ralph` monitor HUD, and a per-theme user-card palette.
 - **[0.4.5]** (2026-06-14) — First-class filesystem make/remove tools.
-- **[0.4.4]** (2026-06-13) — Live subagent status mirroring, always-useful Ctrl+O activity tail, read lineRange crash guard.
-- **[0.4.3]** (2026-06-13) — Readability pass for autopilot, subagent activity, and worked-history review.
 
 See [CHANGELOG.md](CHANGELOG.md) for the full history.
 <!-- CHANGELOG:END -->

package/README.zh.md

@@ -150,11 +150,11 @@ CI 通过 `.github/workflows/npm-publish.yml` 发布 — GitHub 发布 release
 ## 更新日志 (Changelog)
 
 <!-- CHANGELOG:START (auto-generated from CHANGELOG.md — run `bun run changelog:sync`) -->
+- **[0.4.9]** (2026-06-14) — Live-frame width-clamp (content-sized height) replaces the constant-height approach, typed text shows during a running turn, and a docs/AGENTS refresh.
+- **[0.4.8]** (2026-06-14) — Live-frame stability: constant-height live turn, renderer self-heal off-by-one fix, and frame-safe child-stdout sanitizing — no more duplicate model bar or torn escapes.
 - **[0.4.7]** (2026-06-14) — Detached subagents + `subagent` control tool, live shaded in-flight output, registry-driven providers, fuller `read` budget, styled italics in the final report, and `gjc` retired.
 - **[0.4.6]** (2026-06-14) — Width-correct forge cards for CJK/emoji, red borders on failed tool cards, aligned `ooo ralph` monitor HUD, and a per-theme user-card palette.
 - **[0.4.5]** (2026-06-14) — First-class filesystem make/remove tools.
-- **[0.4.4]** (2026-06-13) — Live subagent status mirroring, always-useful Ctrl+O activity tail, read lineRange crash guard.
-- **[0.4.3]** (2026-06-13) — Readability pass for autopilot, subagent activity, and worked-history review.
 
 See [CHANGELOG.md](CHANGELOG.md) for the full history.
 <!-- CHANGELOG:END -->

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "jeo-code",
-  "version": "0.4.7",
+  "version": "0.4.9",
   "description": "Clean, highly optimized AI coding agent using spec-first loop",
   "type": "module",
   "main": "src/cli.ts",

package/src/AGENTS.md

@@ -1,5 +1,5 @@
 <!-- Parent: ../AGENTS.md -->
-<!-- Generated: 2026-06-11 | Updated: 2026-06-11 -->
+<!-- Generated: 2026-06-11 | Updated: 2026-06-14 -->
 
 # src
 
@@ -9,9 +9,12 @@ Core application source code for the `jeo-code` CLI and agent runtime. It orches
 ## Key Files
 | File | Description |
 |------|-------------|
+| `autopilot.ts` | Brief description of purpose |
+| `bun-imports.d.ts` | TypeScript declarations for Bun built-ins and raw module imports |
 | `cli.ts` | The main binary entrypoint |
 | `index.ts` | Library exports |
-| `bun-imports.d.ts` / `md-modules.d.ts` | TypeScript declarations for Bun built-ins and raw module imports |
+| `ledger.ts` | Brief description of purpose |
+| `md-modules.d.ts` | TypeScript declarations for Bun built-ins and raw module imports |
 
 ## Subdirectories
 | Directory | Purpose |

package/src/agent/AGENTS.md

@@ -1,5 +1,5 @@
 <!-- Parent: ../AGENTS.md -->
-<!-- Generated: 2026-06-11 | Updated: 2026-06-11 -->
+<!-- Generated: 2026-06-11 | Updated: 2026-06-14 -->
 
 # agent
 
@@ -9,13 +9,33 @@ The core runtime loop, tool registry, session management, and state persistence
 ## Key Files
 | File | Description |
 |------|-------------|
+| `bash-fixups.ts` | Brief description of purpose |
+| `compaction.ts` | Brief description of purpose |
+| `config-schema.ts` | Brief description of purpose |
+| `context-files.ts` | Brief description of purpose |
+| `engine.ts` | Brief description of purpose |
+| `hooks.ts` | Brief description of purpose |
+| `json.ts` | Brief description of purpose |
 | `loop.ts` | The primary execution loop orchestrating model calls and tool execution |
-| `tools.ts` | Built-in tool definitions (bash, read, write, edit, etc.) |
-| `state.ts` | File-backed state and session persistence (`.jeo/state/`) |
-| `session.ts` | Session context building, compaction, and history management |
+| `memory.ts` | Brief description of purpose |
+| `model-recency.ts` | Brief description of purpose |
+| `output-minimizer.ts` | Brief description of purpose |
+| `output-util.ts` | Brief description of purpose |
 | `plan.ts` | Subagent planning structures and validation |
+| `seed.ts` | Brief description of purpose |
+| `session.ts` | Session context building, compaction, and history management |
+| `state.ts` | File-backed state and session persistence (`.jeo/state/`) |
 | `step-budget.ts` | gjc-style flexible step budgeting: progress-scored extensions, hard cap, fail-fast |
-| `subagents.ts` / `task-tool.ts` | Delegation mechanisms and background execution of task subagents |
+| `subagent-registry.ts` | Brief description of purpose |
+| `subagent-tool.ts` | Brief description of purpose |
+| `subagents.ts` | Brief description of purpose |
+| `task-tool.ts` | Brief description of purpose |
+| `todo-tool.ts` | Brief description of purpose |
+| `tokenizer.ts` | Brief description of purpose |
+| `tool-output.ts` | Brief description of purpose |
+| `tool-registry.ts` | Brief description of purpose |
+| `tools.ts` | Built-in tool definitions (bash, read, write, edit, etc.) |
+| `web-search.ts` | Brief description of purpose |
 
 ## Subdirectories
 | Directory | Purpose |

package/src/agent/dev/AGENTS.md

@@ -1,5 +1,5 @@
-<!-- Parent: ../../AGENTS.md -->
-<!-- Generated: 2026-06-11 | Updated: 2026-06-11 -->
+<!-- Parent: ../AGENTS.md -->
+<!-- Generated: 2026-06-11 | Updated: 2026-06-14 -->
 
 # dev
 
@@ -9,6 +9,11 @@ Developer-specific agent tooling and spec automation.
 ## Key Files
 | File | Description |
 |------|-------------|
+| `advanced-analyzer.ts` | Brief description of purpose |
+| `evolution-bridge.ts` | Brief description of purpose |
+| `evolution-logger.ts` | Brief description of purpose |
+| `self-analysis.ts` | Brief description of purpose |
+| `self-improve.ts` | Brief description of purpose |
 | `spec-automation.ts` | Tools to sync `.specify/` artifacts to ooo seeds |
 
 ## Subdirectories

package/src/ai/AGENTS.md

@@ -1,5 +1,5 @@
 <!-- Parent: ../AGENTS.md -->
-<!-- Generated: 2026-06-11 | Updated: 2026-06-11 -->
+<!-- Generated: 2026-06-11 | Updated: 2026-06-14 -->
 
 # ai
 
@@ -9,9 +9,20 @@ Abstractions for LLM inference, provider management, tool formatting, and token
 ## Key Files
 | File | Description |
 |------|-------------|
-| `registry.ts` | Central registry for available providers and model resolution |
+| `index.ts` | Brief description of purpose |
+| `model-catalog-compat.ts` | Brief description of purpose |
+| `model-catalog.ts` | Brief description of purpose |
+| `model-discovery.ts` | Brief description of purpose |
+| `model-enrich.ts` | Brief description of purpose |
+| `model-manager.ts` | Brief description of purpose |
+| `model-picker.ts` | Brief description of purpose |
+| `model-registry.ts` | Brief description of purpose |
+| `pricing.ts` | Brief description of purpose |
+| `provider-registry.ts` | Brief description of purpose |
+| `provider-status.ts` | Brief description of purpose |
+| `register-providers.ts` | Brief description of purpose |
+| `sse.ts` | Brief description of purpose |
 | `types.ts` | Common interfaces for AI requests, responses, and streams |
-| `format.ts` | Conversion between internal tool schemas and provider-specific formats |
 
 ## Subdirectories
 | Directory | Purpose |

package/src/ai/providers/AGENTS.md

@@ -1,5 +1,5 @@
-<!-- Parent: ../../AGENTS.md -->
-<!-- Generated: 2026-06-11 | Updated: 2026-06-11 -->
+<!-- Parent: ../AGENTS.md -->
+<!-- Generated: 2026-06-11 | Updated: 2026-06-14 -->
 
 # providers
 
@@ -10,10 +10,12 @@ Concrete implementations for various LLM providers, translating generic requests
 | File | Description |
 |------|-------------|
 | `anthropic.ts` | Anthropic Claude integration |
-| `openai.ts` | OpenAI (and Codex backend) integration |
-| `gemini.ts` | Google Gemini (and Cloud Code Assist) integration |
 | `antigravity.ts` | Antigravity desktop-app OAuth client integration |
+| `errors.ts` | Brief description of purpose |
+| `gemini.ts` | Google Gemini (and Cloud Code Assist) integration |
 | `ollama.ts` | Local Ollama integration |
+| `openai-responses.ts` | Brief description of purpose |
+| `openai.ts` | OpenAI (and Codex backend) integration |
 
 ## Subdirectories
 *(None)*

package/src/auth/AGENTS.md

@@ -1,5 +1,5 @@
 <!-- Parent: ../AGENTS.md -->
-<!-- Generated: 2026-06-11 | Updated: 2026-06-11 -->
+<!-- Generated: 2026-06-11 | Updated: 2026-06-14 -->
 
 # auth
 
@@ -9,8 +9,13 @@ Authentication and credential management for OAuth flows and API keys. Ensures s
 ## Key Files
 | File | Description |
 |------|-------------|
-| `store.ts` | Secure credential storage mechanism |
-| `config.ts` | Resolution of keys from environment variables and config files |
+| `callback-server.ts` | Brief description of purpose |
+| `index.ts` | Brief description of purpose |
+| `oauth.ts` | Brief description of purpose |
+| `pkce.ts` | Brief description of purpose |
+| `refresh.ts` | Brief description of purpose |
+| `storage.ts` | Brief description of purpose |
+| `types.ts` | Brief description of purpose |
 
 ## Subdirectories
 | Directory | Purpose |

package/src/auth/flows/AGENTS.md

@@ -1,5 +1,5 @@
-<!-- Parent: ../../AGENTS.md -->
-<!-- Generated: 2026-06-11 | Updated: 2026-06-11 -->
+<!-- Parent: ../AGENTS.md -->
+<!-- Generated: 2026-06-11 | Updated: 2026-06-14 -->
 
 # flows
 
@@ -9,8 +9,12 @@ Specific OAuth flow implementations for various providers.
 ## Key Files
 | File | Description |
 |------|-------------|
-| `oauth.ts` / `pkce.ts` | Generic OAuth and PKCE utilities |
-| `*.ts` | Provider-specific login flows (e.g., anthropic, openai, gemini) |
+| `anthropic.ts` | Brief description of purpose |
+| `antigravity.ts` | Brief description of purpose |
+| `google-project.ts` | Brief description of purpose |
+| `google.ts` | Brief description of purpose |
+| `index.ts` | Brief description of purpose |
+| `openai.ts` | Brief description of purpose |
 
 ## Subdirectories
 *(None)*
@@ -27,6 +31,11 @@ Specific OAuth flow implementations for various providers.
 - Loopback HTTP servers for receiving OAuth callbacks.
 
 ## Dependencies
-*(None)*
+
+### Internal
+{References to other parts of the codebase this depends on}
+
+### External
+{Key external packages/libraries used}
 
 <!-- MANUAL: -->

package/src/cli/AGENTS.md

@@ -1,5 +1,5 @@
 <!-- Parent: ../AGENTS.md -->
-<!-- Generated: 2026-06-11 | Updated: 2026-06-11 -->
+<!-- Generated: 2026-06-11 | Updated: 2026-06-14 -->
 
 # cli
 
@@ -9,8 +9,8 @@ Command-line interface routing, argument parsing, and initialization logic. Defi
 ## Key Files
 | File | Description |
 |------|-------------|
-| `parser.ts` | Argument parsing and flag validation |
-| `router.ts` | Dispatches CLI commands to their respective implementations |
+| `index.ts` | Brief description of purpose |
+| `runner.ts` | Brief description of purpose |
 
 ## Subdirectories
 *(None)*

package/src/commands/AGENTS.md

@@ -1,5 +1,5 @@
 <!-- Parent: ../AGENTS.md -->
-<!-- Generated: 2026-06-11 | Updated: 2026-06-11 -->
+<!-- Generated: 2026-06-11 | Updated: 2026-06-14 -->
 
 # commands
 
@@ -9,10 +9,28 @@ Concrete implementations of `jeo` subcommands (e.g., launch, setup, team, ultrag
 ## Key Files
 | File | Description |
 |------|-------------|
+| `approve.ts` | Brief description of purpose |
+| `auth.ts` | Brief description of purpose |
+| `chat.ts` | Brief description of purpose |
+| `deep-interview.ts` | Brief description of purpose |
+| `doctor.ts` | Brief description of purpose |
+| `evolve-core.ts` | Brief description of purpose |
+| `evolve.ts` | Brief description of purpose |
+| `export.ts` | Brief description of purpose |
 | `launch.ts` | The primary interactive/one-shot execution command |
+| `mcp.ts` | Brief description of purpose |
+| `ooo-seed.ts` | Brief description of purpose |
+| `ralplan.ts` | Brief description of purpose |
+| `resume.ts` | Brief description of purpose |
+| `session.ts` | Brief description of purpose |
+| `setup-helpers.ts` | Brief description of purpose |
 | `setup.ts` | Guided configuration command |
+| `skills.ts` | Brief description of purpose |
+| `state.ts` | Brief description of purpose |
+| `status.ts` | Brief description of purpose |
 | `team.ts` | Multi-agent coordination command |
-| `status.ts` / `update.ts` | Inspection and maintenance commands |
+| `ultragoal.ts` | Brief description of purpose |
+| `update.ts` | Brief description of purpose |
 
 ## Subdirectories
 *(None)*

package/src/commands/launch.ts

@@ -1418,13 +1418,13 @@ export async function runLaunchCommand(args: string[]): Promise<void> {
               return;
             }
           }
-          // Mid-turn additional input is committed (and shown) ONLY on Enter (above):
-          // the running turn does NOT echo half-typed text per keystroke. Captured
-          // printable input accumulates silently in the draft buffer and surfaces as a
-          // `user` card the moment Enter lifts it into the steering inbox (or folds into
-          // the next prompt if the turn ends first). JEO_LIVE_DRAFT=1 restores the
-          // legacy live per-keystroke echo in the input box.
-          if (captured && jeoEnv("LIVE_DRAFT") === "1") {
+          // Mid-turn typing is echoed LIVE into the input box so the field never looks
+          // frozen/disabled while a turn runs — the typed text IS captured, so it must be
+          // visible. This updates the in-frame input box via the differential renderer
+          // only; it never leaks into scrollback/history (native readline echo stays
+          // suppressed for the whole turn). On Enter the draft is lifted into the steering
+          // inbox and surfaces as a `user` card (above). JEO_NO_LIVE_DRAFT=1 opts out.
+          if (captured && jeoEnv("NO_LIVE_DRAFT") !== "1") {
             tui.setLivePromptInput(queueBusySnapshot?.().text ?? "");
           }
         },

package/src/mcp/AGENTS.md

@@ -1,5 +1,5 @@
 <!-- Parent: ../AGENTS.md -->
-<!-- Generated: 2026-06-11 | Updated: 2026-06-11 -->
+<!-- Generated: 2026-06-11 | Updated: 2026-06-14 -->
 
 # mcp
 
@@ -9,8 +9,10 @@ Model Context Protocol (MCP) integration, allowing `jeo-code` to expose tools or
 ## Key Files
 | File | Description |
 |------|-------------|
+| `index.ts` | Brief description of purpose |
+| `protocol.ts` | Brief description of purpose |
 | `server.ts` | MCP server implementation (if exposing tools) |
-| `client.ts` | MCP client implementation (for consuming external tools) |
+| `tools.ts` | Brief description of purpose |
 
 ## Subdirectories
 *(None)*

package/src/prompts/AGENTS.md

@@ -1,5 +1,5 @@
 <!-- Parent: ../AGENTS.md -->
-<!-- Generated: 2026-06-11 | Updated: 2026-06-11 -->
+<!-- Generated: 2026-06-11 | Updated: 2026-06-14 -->
 
 # prompts
 
@@ -7,9 +7,7 @@
 Bundled system prompts, role definitions, and workflow skill documentation for the agent.
 
 ## Key Files
-| File | Description |
-|------|-------------|
-| `system.md` | Core system prompt template |
+*(None)*
 
 ## Subdirectories
 | Directory | Purpose |

package/src/prompts/agents/AGENTS.md

@@ -1,5 +1,5 @@
-<!-- Parent: ../../AGENTS.md -->
-<!-- Generated: 2026-06-11 | Updated: 2026-06-11 -->
+<!-- Parent: ../AGENTS.md -->
+<!-- Generated: 2026-06-11 | Updated: 2026-06-14 -->
 
 # agents
 
@@ -9,10 +9,10 @@ Role-specific system prompts defining the behavior of subagents (executor, plann
 ## Key Files
 | File | Description |
 |------|-------------|
-| `executor.md` | Mutating agent for executing bounded tasks |
-| `planner.md` | Read-only agent for sequencing tasks |
 | `architect.md` | Read-only agent for structural review |
 | `critic.md` | Read-only agent for plan critique |
+| `executor.md` | Mutating agent for executing bounded tasks |
+| `planner.md` | Read-only agent for sequencing tasks |
 
 ## Subdirectories
 *(None)*
@@ -30,6 +30,11 @@ Role-specific system prompts defining the behavior of subagents (executor, plann
 *(None)*
 
 ## Dependencies
-*(None)*
+
+### Internal
+{References to other parts of the codebase this depends on}
+
+### External
+{Key external packages/libraries used}
 
 <!-- MANUAL: -->

package/src/prompts/skills/AGENTS.md

@@ -1,5 +1,5 @@
-<!-- Parent: ../../AGENTS.md -->
-<!-- Generated: 2026-06-11 | Updated: 2026-06-11 -->
+<!-- Parent: ../AGENTS.md -->
+<!-- Generated: 2026-06-11 | Updated: 2026-06-14 -->
 
 # skills
 
@@ -7,16 +7,16 @@
 Bundled workflow skills offering specialized, multi-step procedures for the agent.
 
 ## Key Files
-*(See subdirectories)*
+*(None)*
 
 ## Subdirectories
 | Directory | Purpose |
 |-----------|---------|
+| `deep-dive/` | What it contains (see `deep-dive/AGENTS.md`) |
 | `deep-interview/` | Socratic requirements gathering |
 | `ralplan/` | Spec-first planning and consensus building |
 | `team/` | Coordinated multi-agent execution |
 | `ultragoal/` | Durable verification and ledger tracking |
-| `gjc/` | Legacy parity configurations |
 
 ## For AI Agents
 
@@ -31,6 +31,11 @@ Bundled workflow skills offering specialized, multi-step procedures for the agen
 *(None)*
 
 ## Dependencies
-*(None)*
+
+### Internal
+{References to other parts of the codebase this depends on}
+
+### External
+{Key external packages/libraries used}
 
 <!-- MANUAL: -->

package/src/prompts/skills/deep-dive/AGENTS.md

@@ -1,5 +1,5 @@
 <!-- Parent: ../AGENTS.md -->
-<!-- Generated: 2026-06-11 | Updated: 2026-06-11 -->
+<!-- Generated: 2026-06-11 | Updated: 2026-06-14 -->
 
 # deep-dive
 
@@ -26,6 +26,11 @@ Bundled SKILL.md for the `deep-dive` workflow.
 *(None)*
 
 ## Dependencies
-*(None)*
+
+### Internal
+{References to other parts of the codebase this depends on}
+
+### External
+{Key external packages/libraries used}
 
 <!-- MANUAL: -->

package/src/prompts/skills/deep-interview/AGENTS.md

@@ -1,5 +1,5 @@
 <!-- Parent: ../AGENTS.md -->
-<!-- Generated: 2026-06-11 | Updated: 2026-06-11 -->
+<!-- Generated: 2026-06-11 | Updated: 2026-06-14 -->
 
 # deep-interview
 
@@ -26,6 +26,11 @@ Bundled SKILL.md for the `deep-interview` workflow.
 *(None)*
 
 ## Dependencies
-*(None)*
+
+### Internal
+{References to other parts of the codebase this depends on}
+
+### External
+{Key external packages/libraries used}
 
 <!-- MANUAL: -->

package/src/prompts/skills/ralplan/AGENTS.md

@@ -1,5 +1,5 @@
 <!-- Parent: ../AGENTS.md -->
-<!-- Generated: 2026-06-11 | Updated: 2026-06-11 -->
+<!-- Generated: 2026-06-11 | Updated: 2026-06-14 -->
 
 # ralplan
 
@@ -26,6 +26,11 @@ Bundled SKILL.md for the `ralplan` workflow.
 *(None)*
 
 ## Dependencies
-*(None)*
+
+### Internal
+{References to other parts of the codebase this depends on}
+
+### External
+{Key external packages/libraries used}
 
 <!-- MANUAL: -->

package/src/prompts/skills/team/AGENTS.md

@@ -1,5 +1,5 @@
 <!-- Parent: ../AGENTS.md -->
-<!-- Generated: 2026-06-11 | Updated: 2026-06-11 -->
+<!-- Generated: 2026-06-11 | Updated: 2026-06-14 -->
 
 # team
 
@@ -26,6 +26,11 @@ Bundled SKILL.md for the `team` workflow.
 *(None)*
 
 ## Dependencies
-*(None)*
+
+### Internal
+{References to other parts of the codebase this depends on}
+
+### External
+{Key external packages/libraries used}
 
 <!-- MANUAL: -->

package/src/prompts/skills/ultragoal/AGENTS.md

@@ -1,5 +1,5 @@
 <!-- Parent: ../AGENTS.md -->
-<!-- Generated: 2026-06-11 | Updated: 2026-06-11 -->
+<!-- Generated: 2026-06-11 | Updated: 2026-06-14 -->
 
 # ultragoal
 
@@ -26,6 +26,11 @@ Bundled SKILL.md for the `ultragoal` workflow.
 *(None)*
 
 ## Dependencies
-*(None)*
+
+### Internal
+{References to other parts of the codebase this depends on}
+
+### External
+{Key external packages/libraries used}
 
 <!-- MANUAL: -->

package/src/skills/AGENTS.md

@@ -1,5 +1,5 @@
 <!-- Parent: ../AGENTS.md -->
-<!-- Generated: 2026-06-11 | Updated: 2026-06-11 -->
+<!-- Generated: 2026-06-11 | Updated: 2026-06-14 -->
 
 # skills
 
@@ -9,8 +9,7 @@ Framework for discovering, loading, and executing workflow skills (both bundled
 ## Key Files
 | File | Description |
 |------|-------------|
-| `registry.ts` | Discovers and catalogs available skills |
-| `executor.ts` | Mechanism to inject skill prompts into the active session |
+| `catalog.ts` | Brief description of purpose |
 
 ## Subdirectories
 *(None)*

package/src/tui/AGENTS.md

@@ -1,5 +1,5 @@
 <!-- Parent: ../AGENTS.md -->
-<!-- Generated: 2026-06-11 | Updated: 2026-06-11 -->
+<!-- Generated: 2026-06-11 | Updated: 2026-06-14 -->
 
 # tui
 
@@ -10,6 +10,7 @@ Terminal User Interface logic, layout systems, and the differential renderer. Re
 | File | Description |
 |------|-------------|
 | `app.ts` | High-level TUI orchestrator (`LaunchTui`) |
+| `index.ts` | Brief description of purpose |
 | `renderer.ts` | The differential, atomic terminal renderer (handles scrollback, resizing, DECSET 2026) |
 | `terminal.ts` | Low-level ANSI escape codes and terminal size utilities |
 

package/src/tui/app.ts

@@ -30,7 +30,7 @@ import { costForUsage, formatCost } from "../ai/pricing";
 import { renderMarkdownTables } from "./components/markdown-table";
  
 import { stripMarkdown, renderMarkdownAnsi } from "./components/markdown-text";
-import { visibleWidth, wrapTextWithAnsi } from "./components/width";
+import { visibleWidth, wrapTextWithAnsi, truncateToWidth, sanitizeForFrame } from "./components/width";
 import { categoryBadge } from "./components/category-index";
 import { formatStepTimeline, stepsFromTools, formatStepHeader, formatStepTimelineCompact, type StepState } from "./components/step-timeline";
 import { formatHintBar } from "./components/hints";
@@ -428,7 +428,10 @@ export class LaunchTui {
       },
       onToolProgress: (_tool, partial) => {
         if (this.finished) return;
-        this.liveToolOutput = partial;
+        // Sanitize raw child stdout (CR / EL / cursor-move escapes) before it enters the
+        // frame — unsanitized control bytes tore the renderer's next \x1b[2K (literal "2K")
+        // and hijacked the cursor, corrupting the live frame.
+        this.liveToolOutput = sanitizeForFrame(partial);
         if (Date.now() - this.lastStreamDraw >= 100) {
           this.lastStreamDraw = Date.now();
           this.draw();
@@ -551,10 +554,6 @@ export class LaunchTui {
     });
   }
 
-  private renderLiveUserQueryCard(cols: number): string[] {
-    return this.renderUserCard(this.livePromptInput, cols);
-  }
-
   /** Render a `user`-labeled query card (orange "user" header over a filled box).
    *  Shared by the live next-prompt draft and the mid-turn steering flush. */
   private renderUserCard(rawText: string, cols: number): string[] {
@@ -1119,12 +1118,16 @@ export class LaunchTui {
         .split("\n")
         .flatMap(l => wrapTextWithAnsi(l, wrapW))
         .filter(l => l.length > 0);
-      const shown = wrapped.slice(-6); // bottom-anchored tail of the live trace
-      if (shown.length) {
-        tail.push(dim(`${this.unicode ? "│" : "|"} thinking`));
-        for (const l of shown) tail.push(dim(`  ${l}`));
-        tail.push("");
-      }
+      // FIXED reserved height (bottom-anchored, blank-padded at top): once present the
+      // block's row count is CONSTANT, so streaming content never changes the frame
+      // height. The per-100ms height thrash that desynced the differential renderer
+      // (duplicate model bar) is gone; height now toggles only at lifecycle boundaries.
+      const ROWS = 6;
+      const shown = wrapped.slice(-ROWS);
+      tail.push(dim(`${this.unicode ? "│" : "|"} thinking`));
+      for (let k = 0; k < ROWS - shown.length; k++) tail.push("");
+      for (const l of shown) tail.push(dim(`  ${l}`));
+      tail.push("");
     }
 
     // Live tool output (gjc-style streaming bash stdout): while a tool runs, its
@@ -1136,12 +1139,14 @@ export class LaunchTui {
         .split("\n")
         .flatMap(l => wrapTextWithAnsi(l, wrapW))
         .filter(l => l.length > 0);
-      const shown = wrapped.slice(-8); // bottom-anchored tail of the live output
-      if (shown.length) {
-        tail.push(dim(`${this.unicode ? "│" : "|"} output`));
-        for (const l of shown) tail.push(dim(`  ${l}`));
-        tail.push("");
-      }
+      // FIXED reserved height (see thinking block): constant rows while a tool streams,
+      // so cumulative stdout growth does not thrash the frame height.
+      const ROWS = 8;
+      const shown = wrapped.slice(-ROWS);
+      tail.push(dim(`${this.unicode ? "│" : "|"} output`));
+      for (let k = 0; k < ROWS - shown.length; k++) tail.push("");
+      for (const l of shown) tail.push(dim(`  ${l}`));
+      tail.push("");
     }
 
     // Live status field: unboxed thinking line + compact metrics row. The model's
@@ -1177,12 +1182,6 @@ export class LaunchTui {
       }));
     }
 
-    // User text typed while the model is still thinking is surfaced as a live
-    // pending-user card (same query input underneath, no hidden auto-executing queue).
-    if (this.livePromptInput.trim()) {
-      tail.push("");
-      tail.push(...this.renderLiveUserQueryCard(cols));
-    }
 
     // Agent task plan (the `todo` tool) as a Todos checklist.
     if (args.planLines.length) {
@@ -1307,7 +1306,11 @@ export class LaunchTui {
     // model bar), no outer border, no mascot art — completed work lives in scrollback.
     if (fit && this.inline) {
       const inlineFrame = this.composeInlineFrame({ cols, rows, stepNow, elapsedMs, idx, isThinking, planLines });
-      this.renderer.render(inlineFrame.slice(0, rows));
+      // Screen-safety: every rendered line is width-clamped to `cols` so a long line
+      // (e.g. the model bar with a deep cwd) cannot soft-wrap into a second physical row
+      // and desync the differential renderer's 1-line=1-row accounting. Frame height stays
+      // content-sized so completed cards remain visible in scrollback above the live frame.
+      this.renderer.render(inlineFrame.slice(0, rows).map(l => truncateToWidth(l, cols)));
       return;
     }
 
@@ -1371,10 +1374,6 @@ export class LaunchTui {
         bottom.push(`  ${categoryBadge("status", { color: this.theme.color })} ${msg}${guardBadge}`);
       }
     }
-    if (fit && this.livePromptInput.trim()) {
-      bottom.push("");
-      bottom.push(...this.renderLiveUserQueryCard(innerWidth));
-    }
     // TTY only: keep the same query input box visible above the footer while the
     // turn is running; typed text edits the next-prompt draft, not a side queue.
     if (fit) {
@@ -1492,6 +1491,6 @@ export class LaunchTui {
     if (fit) {
       frame = frame.slice(0, rows);
     }
-    this.renderer.render(frame);
+    this.renderer.render(frame.map(l => truncateToWidth(l, cols)));
   }
 }

package/src/tui/components/AGENTS.md

@@ -1,5 +1,5 @@
-<!-- Parent: ../../AGENTS.md -->
-<!-- Generated: 2026-06-11 | Updated: 2026-06-11 -->
+<!-- Parent: ../AGENTS.md -->
+<!-- Generated: 2026-06-11 | Updated: 2026-06-14 -->
 
 # components
 
@@ -9,11 +9,44 @@ Reusable UI widgets and layout primitives for the terminal interface.
 ## Key Files
 | File | Description |
 |------|-------------|
+| `ascii-art.ts` | Cellular evolution graphics |
+| `autocomplete.ts` | Brief description of purpose |
+| `autopilot-status.ts` | Brief description of purpose |
+| `capability.ts` | Brief description of purpose |
+| `category-index.ts` | Brief description of purpose |
+| `code-view.ts` | Brief description of purpose |
+| `color.ts` | Brief description of purpose |
+| `config-panel.ts` | Brief description of purpose |
+| `duration.ts` | Brief description of purpose |
+| `evolution.ts` | Brief description of purpose |
+| `footer.ts` | Brief description of purpose |
 | `forge.ts` | Formats the boxed tool execution outputs |
-| `status.ts` | The `[STEP]` / `[STATUS]` / `[TOOL]` HUD lines |
-| `section.ts` | Shadcn-inspired card layout and spacing tokens |
+| `hints.ts` | Brief description of purpose |
+| `hud.ts` | Brief description of purpose |
+| `index.ts` | Brief description of purpose |
+| `input-box.ts` | Brief description of purpose |
 | `layout.ts` | Low-level padding, boxing, and alignment math |
-| `ascii-art.ts` | Cellular evolution graphics |
+| `live-model-picker.ts` | Brief description of purpose |
+| `markdown-table.ts` | Brief description of purpose |
+| `markdown-text.ts` | Brief description of purpose |
+| `meter.ts` | Brief description of purpose |
+| `model-picker.ts` | Brief description of purpose |
+| `provider-picker.ts` | Brief description of purpose |
+| `section.ts` | Shadcn-inspired card layout and spacing tokens |
+| `select-list.ts` | Brief description of purpose |
+| `skill-picker.ts` | Brief description of purpose |
+| `slash.ts` | Brief description of purpose |
+| `spinner.ts` | Brief description of purpose |
+| `status.ts` | The `[STEP]` / `[STATUS]` / `[TOOL]` HUD lines |
+| `step-timeline.ts` | Brief description of purpose |
+| `stream.ts` | Brief description of purpose |
+| `themes.ts` | Brief description of purpose |
+| `todo-card.ts` | Brief description of purpose |
+| `tool-list.ts` | Brief description of purpose |
+| `transcript.ts` | Brief description of purpose |
+| `update-box.ts` | Brief description of purpose |
+| `welcome.ts` | Brief description of purpose |
+| `width.ts` | Brief description of purpose |
 
 ## Subdirectories
 *(None)*

package/src/tui/components/width.ts

@@ -134,6 +134,57 @@ export function truncateToWidth(s: string, cols: number): string {
   return out;
 }
 
+/**
+ * Strip control bytes that would corrupt the live differential frame. KEEPS SGR color
+ * escapes (`\x1b[…m`); DROPS every other CSI (cursor moves, EL/ED erase, etc.), OSC
+ * sequences, other escapes, and bare C0/C1 control bytes except tab/newline; and DROPS
+ * an INCOMPLETE trailing escape (a chunk that ends mid-sequence) so it can never eat the
+ * next line's `\x1b[2K`. Used to sanitize raw child stdout (e.g. a streaming `bun test`
+ * with `\r\x1b[2K` progress lines) before it enters the frame — the torn-escape /
+ * cursor-hijack class of screen corruption.
+ */
+export function sanitizeForFrame(s: string): string {
+  if (!s.includes("\x1b") && !/[\x00-\x08\x0b-\x1f\x7f]/.test(s)) return s; // fast path
+  let out = "";
+  let i = 0;
+  const n = s.length;
+  while (i < n) {
+    const ch = s[i]!;
+    if (ch === "\x1b") {
+      if (s[i + 1] === "[") {
+        // CSI: ESC [ params (0-9;:?<>= space) final (@-~)
+        let j = i + 2;
+        while (j < n && /[0-9;:?<>= ]/.test(s[j]!)) j++;
+        if (j < n && /[@-~]/.test(s[j]!)) {
+          const seq = s.slice(i, j + 1);
+          if (seq.endsWith("m")) out += seq; // keep SGR color, drop all other CSI
+          i = j + 1;
+          continue;
+        }
+        break; // incomplete CSI at the chunk tail → drop the rest
+      }
+      if (s[i + 1] === "]") {
+        // OSC: ESC ] … (BEL | ST = ESC \)
+        let j = i + 2;
+        while (j < n && s[j] !== "\x07" && !(s[j] === "\x1b" && s[j + 1] === "\\")) j++;
+        if (j >= n) break; // incomplete OSC → drop
+        i = s[j] === "\x07" ? j + 1 : j + 2;
+        continue;
+      }
+      i += i + 1 < n ? 2 : 1; // other ESC x → drop ESC (+ its single byte)
+      continue;
+    }
+    const code = ch.charCodeAt(0);
+    if ((code < 0x20 && code !== 0x09 && code !== 0x0a) || code === 0x7f) {
+      i++; // strip bare control bytes (CR/BS/…), keep tab + newline
+      continue;
+    }
+    out += ch;
+    i++;
+  }
+  return out;
+}
+
 /**
  * Hard-wrap text to `cols` display columns, breaking long words and preserving
  * existing newlines. SGR-aware (escapes don't consume width). Returns the wrapped

package/src/tui/monitoring/AGENTS.md

@@ -1,5 +1,5 @@
-<!-- Parent: ../../AGENTS.md -->
-<!-- Generated: 2026-06-11 | Updated: 2026-06-11 -->
+<!-- Parent: ../AGENTS.md -->
+<!-- Generated: 2026-06-11 | Updated: 2026-06-14 -->
 
 # monitoring
 
@@ -26,6 +26,11 @@ Specialized HUD views and sovereign monitoring interfaces.
 *(None)*
 
 ## Dependencies
-*(None)*
+
+### Internal
+{References to other parts of the codebase this depends on}
+
+### External
+{Key external packages/libraries used}
 
 <!-- MANUAL: -->

package/src/tui/renderer.ts

@@ -51,17 +51,22 @@ export class Renderer {
     this.prevCols = currentCols;
 
     const next = lines.map(line => truncate(line, currentCols));
+    // Rows physically occupied by the prior frame — or recorded by reset() when the
+    // baseline was dropped WITHOUT clearing the screen. The diff below EL-clears any
+    // of these that the new (possibly shorter) frame does not cover, and the reserve
+    // block below uses it so a post-reset repaint does not spuriously re-scroll.
+    const occupied = Math.max(this.prev.length, this.coverRows);
     const maxLen = Math.max(this.prev.length, next.length, this.coverRows);
     this.coverRows = 0;
     let cursorRow = 0;
     let out = "";
 
-    if (this.reserve && next.length > this.prev.length && next.length <= Math.max(1, size().rows)) {
+    if (this.reserve && next.length > occupied && next.length <= Math.max(1, size().rows)) {
       // The cursor rests on the frame's first row (the anchor). Walk to the last
       // currently-occupied row, emit one newline per missing row (scrolling the
       // viewport when at the bottom margin), then hop back up to the — possibly
       // shifted — anchor so the diff below paints at stable relative positions.
-      const have = Math.max(this.prev.length, 1);
+      const have = Math.max(occupied, 1);
       out += cursorDown(have - 1) + "\n".repeat(next.length - have) + cursorUp(next.length - 1) + toColumn(1);
     }
 
@@ -92,16 +97,17 @@ export class Renderer {
     }
     out += toColumn(1);
 
-    // Close the synchronized update opened by insertAbove() now that the full
-    // repaint is in the same buffered stream — the terminal presents the overwritten
-    // first row(s), the flushed ledger line, and the repainted frame as ONE atomic update.
-    if (this.syncOpen) {
-      out += END_SYNC;
-      this.syncOpen = false;
-    }
-
+    // Atomic present (DECSET-2026 synchronized update): wrap the WHOLE repaint so the
+    // terminal never shows a half-painted frame — no torn row, no transient duplicate
+    // bar a mid-repaint snapshot could catch. An insertAbove() may have already opened
+    // the update (syncOpen); otherwise this render opens its own. Exactly one BSU/ESU
+    // pair is emitted per write.
     if (out.length > 0) {
-      this.write(out);
+      this.write((this.syncOpen ? "" : BEGIN_SYNC) + out + END_SYNC);
+      this.syncOpen = false;
+    } else if (this.syncOpen) {
+      this.write(END_SYNC);
+      this.syncOpen = false;
     }
 
     this.prev = next;
@@ -128,16 +134,12 @@ export class Renderer {
       .map((line, i, arr) => (i === arr.length - 1 && line === "" ? "" : toColumn(1) + clearLine() + line))
       .join("\n");
     let out = BEGIN_SYNC + body;
-    // Eagerly EL-clear the old frame rows the inserted block did NOT cover, then hop
-    // back to the row right below the insert (where the next render() anchors).
-    // The geometry is provably safe HERE: when stale > 0 the body write never hit
-    // the bottom margin (the old frame fit on screen and the insert is shorter), so
-    // every stale row exists and cursor-down cannot clamp. Deferring this clear to
-    // the next render() via coverRows walked PAST the bottom margin, where the
-    // clamped cursor-down desynced the row bookkeeping — each subsequent frame then
-    // painted one row higher, devouring the flushed scrollback content above (the
-    // "truncated card" corruption).
-    const stale = this.prev.length - written;
+    // EL-clear the old frame rows the inserted block did NOT cover, then hop back to the
+    // row right below the insert (where the next render() anchors). occupied = max(prev,
+    // coverRows) matches the reserve block so a reset()->insertAbove() ordering still
+    // clears the old frame's lower rows (the off-by-one that duplicated the model bar).
+    const occupied = Math.max(this.prev.length, this.coverRows);
+    const stale = occupied - written;
     if (stale > 0) {
       for (let i = 0; i < stale; i++) {
         out += toColumn(1) + clearLine() + (i < stale - 1 ? cursorDown(1) : "");
@@ -146,6 +148,7 @@ export class Renderer {
     }
     this.write(out);
     this.prev = [];
+    this.coverRows = 0;
   }
 
   /** Clear the live frame. Inline (reserve) mode walks the known frame rows with
@@ -174,6 +177,19 @@ export class Renderer {
   }
 
   reset(): void {
+    // Drop the diff baseline so the next render() repaints every line — but REMEMBER
+    // how many rows are physically on screen so that repaint also EL-clears any the
+    // new (possibly shorter) frame doesn't cover. Without this, a self-heal reset on a
+    // frame that just shrank left stale rows behind (duplicate model bars / orphaned
+    // borders — the live-analysis screen corruption).
+    this.coverRows = Math.max(this.coverRows, this.prev.length);
     this.prev = [];
+    // Close any synchronized update opened by a preceding insertAbove() so a reset()
+    // landing between insertAbove() and the next render() cannot strand an open BSU
+    // window (which times out ~150ms later and flashes a partial frame).
+    if (this.syncOpen) {
+      this.write(END_SYNC);
+      this.syncOpen = false;
+    }
   }
 }

package/src/util/AGENTS.md

@@ -1,5 +1,5 @@
 <!-- Parent: ../AGENTS.md -->
-<!-- Generated: 2026-06-11 | Updated: 2026-06-11 -->
+<!-- Generated: 2026-06-11 | Updated: 2026-06-14 -->
 
 # util
 
@@ -9,9 +9,11 @@ General utilities, helper functions, and shared types used across the applicatio
 ## Key Files
 | File | Description |
 |------|-------------|
-| `update-check.ts` | Async check for newer npm versions |
-| `retry.ts` | Rate-limit backoff and generic retry mechanisms |
+| `clipboard-image.ts` | Brief description of purpose |
+| `env.ts` | Brief description of purpose |
 | `provider-error.ts` | Error normalization helpers |
+| `retry.ts` | Rate-limit backoff and generic retry mechanisms |
+| `update-check.ts` | Async check for newer npm versions |
 
 ## Subdirectories
 *(None)*