@oh-my-pi/pi-coding-agent 6.9.69 → 8.0.0

package/CHANGELOG.md

@@ -2,6 +2,150 @@
 
 ## [Unreleased]
 
+## [8.0.0] - 2026-01-23
+### Added
+
+- Added antigravity provider support for image generation with Google Cloud authentication
+- Added support for google-antigravity API credentials in model registry
+- Added antigravity-specific request handling with SSE streaming
+- Added projectId parameter to antigravity credentials parsing
+- Added antigravity provider to preferred image provider selection
+- Added list-limit utility for consistent result limiting across tools
+- Added output-utils module with tail buffer and artifact allocation helpers
+- Added tool-result module for standardized tool result construction
+- Added truncation summary options to OutputMetaBuilder for better output tracking
+- Added artifact storage system for truncated tool outputs with artifact:// URL protocol
+- Added structured output metadata system with fluent OutputMetaBuilder for consistent notices
+- Added standardized tool error types (ToolError, MultiError, ToolAbortError) for better error handling
+- Added internal URL routing system with protocol handlers:
+- `agent://<id>` - access agent output artifacts
+- `agent://<id>/<path>` and `agent://<id>?q=<query>` - JSON extraction from agent outputs
+- `skill://<name>` and `skill://<name>/<path>` - read skill files and relative paths
+- `rule://<name>` - read rule content
+- URL resolution includes filesystem path in output for bash/python interop
+- Added fetch tool for URL content retrieval with enhanced processing capabilities
+- Added `isolated` option to task tool for git worktree execution with automatic patch generation and application
+- Added format-prompts script to standardize prompt file formatting
+
+### Changed
+
+- Updated default line limit from 4000 to 3000 lines for output truncation
+- Reordered truncation notice to show offset continuation before artifact reference
+- Applied meta notice wrapper to all tools in createTools function
+- Updated test expectations to reflect new 3000 line limit
+- Removed output tool from schema validation test list
+- Replaced inline output truncation notices with structured metadata system across all tools
+- Updated bash, python, and ssh executors to track detailed output statistics (total lines/bytes vs output lines/bytes)
+- Modified artifact storage to use pre-allocated paths instead of inline file writing
+- Changed message format to use meta field instead of fullOutputPath for truncation information
+- Updated interactive components to display truncation metadata from structured format
+- Standardized tool result building with new ToolResultBuilder for consistent metadata handling
+- Simplified Python gateway coordination by removing reference counting and client tracking
+- Updated Python gateway to use global shared instance instead of per-process coordination
+- Modified Python kernel initialization to set working directory and environment per kernel
+- Updated interactive status display to show Python and venv paths instead of client count
+- Changed system prompt to clarify CHECKPOINT step 0 timing
+- Updated Python environment warming to use await instead of void for proper error handling
+- Updated interactive mode shutdown to use postmortem.quit instead of process.exit
+- Updated bash tool documentation to clarify specialized tool usage
+- Updated task tool documentation to escape placeholder syntax in examples
+- Updated Python environment warming to use await instead of void for proper error handling
+- Updated interactive mode shutdown to use postmortem.quit instead of process.exit
+- Updated bash tool documentation to clarify specialized tool usage
+- Updated task tool documentation to escape placeholder syntax in examples
+- Updated all tools to use structured metadata instead of inline notices for truncation, limits, and diagnostics
+- Replaced manual error formatting with ToolError.render() and standardized error handling
+- Enhanced bash and python executors to save full output as artifacts when truncated
+- Improved abort signal handling across all tools with consistent ToolAbortError
+- Renamed task parameter from `vars` to `args` throughout task tool interface and updated template rendering to support built-in `{{id}}` and `{{description}}` placeholders
+- Simplified todo-write tool by removing active_form parameter, using single content field for task descriptions
+- Updated system prompt structure with `<important>` and `<avoid>` tags, clearer critical sections, and standardized whitespace handling
+- Renamed web_fetch tool to fetch and removed internal URL handling (use read tool instead)
+- Standardized tool parameter names from camelCase to snake_case across edit, grep, python, and todo-write tools
+- Unified timeout parameters across all tools with auto-conversion from milliseconds and reasonable clamping (1s-3600s for bash/ssh, 1s-600s for python/gemini-image)
+- Simplified web-search tool by removing advanced parameters (`max_tokens`, `model`, `search_domain_filter`, `search_context_size`, `return_related_questions`) and using `recency` instead of `search_recency_filter`
+- Restructured tool documentation with standardized `<instruction>`, `<output>`, `<critical>`, and `<avoid>` sections across all 18 tools
+- Updated find tool to always sort results by modification time
+- Updated bash prompt to use `cwd` parameter instead of `workdir`
+- Improved output truncation limits: bash to 50KB/2000 lines, python to 100KB
+- Removed model parameter from task and gemini-image tools to use session/provider defaults
+- Improved MCP tool name handling with explicit server and tool name properties
+- Marked read tool as non-abortable to improve performance
+- Converted dynamic imports to static imports in installer and exa tools
+
+### Removed
+
+- Removed output tool (replaced by `agent://` URLs via read tool)
+- Removed web_fetch tool (replaced by fetch tool)
+
+### Fixed
+
+- Fixed Python kernel environment initialization for external and shared gateways
+- Fixed gateway status reporting to include Python and virtual environment paths
+- Fixed inconsistent error formatting across tools by standardizing on ToolError types
+- Fixed timeout parameter handling to auto-convert milliseconds to seconds and clamp to reasonable ranges
+- Fixed whitespace formatting in json-query.ts comment
+- Fixed interactive shutdown to await postmortem cleanup so Python kernel gateways are terminated
+- Fixed shared Python gateway reuse across working directories by initializing kernel cwd and env per kernel
+- Fixed Python gateway coordination to use a single global gateway without ref counting
+
+## [7.0.0] - 2026-01-21
+### Added
+
+- Added usage report deduplication to prevent duplicate account entries
+- Added debug logging for usage fetch operations to aid diagnostics
+- Added provider sorting in usage display by total usage amount
+- Added `isolated` parameter to task tool for running each task in separate git worktrees
+- Added git worktree management for isolated task execution with patch generation
+- Added patch application system that applies changes only when all patches are valid
+- Added working directory information to environment info display
+- Added `/usage` command to display provider usage and limits
+- Added support for multiple usage providers beyond Codex
+- Added usage report caching with configurable TTL
+- Added visual usage bars and account aggregation in usage display
+- Added `fetchUsageReports()` method to agent session
+- Added `output()` function to read task/agent outputs by ID with support for multiple formats and queries
+- Added session file support to Python executor for accessing task outputs
+- Added support for jq-like queries when reading JSON outputs
+- Added offset and limit parameters for reading specific line ranges from outputs
+- Added "." and "c" shortcuts to continue agent without sending visible message
+- Added debug logging for usage fetch results to aid /usage diagnostics
+
+### Changed
+
+- Updated discoverSkills function to return object with skills property
+- Enhanced usage report merging to combine limits and metadata from duplicate accounts
+- Improved OAuth credential handling to preserve existing fields when updating
+- Removed cd function from Python prelude to encourage using cwd parameter
+- Updated task tool to generate and apply patches when running in isolated mode
+- Enhanced task tool rendering to display isolated execution status and patch paths
+- Updated system prompt structure and formatting for better readability
+- Reorganized tool hierarchy and discipline sections
+- Added parallel work guidance for task-based workflows
+- Enhanced verification and integration methodology sections
+- Updated skills and rules formatting for cleaner presentation
+- Added stronger emphasis on completeness and quality standards
+- Refactored usage tracking from Codex-specific to generic provider system
+- Updated usage limit detection to work with multiple provider APIs
+- Changed usage cache to use persistent storage instead of in-memory only
+- Limited diagnostic messages to 50 items to prevent overwhelming output when processing files with many issues
+- Changed `/dump` command to include complete agent context: system prompt, model config, available tools with schemas, and all message types (bash/python executions, custom messages, branch summaries, compaction summaries, file mentions)
+- Changed `/dump` format to use YAML instead of JSON for tool schemas and arguments (more readable)
+
+### Fixed
+
+- Fixed TypeScript error in bash executor by properly typing caught exception
+- Fixed usage display ordering to show providers with lowest usage first
+- Fixed task tool result rendering to show fallback text when no results are available
+- Fixed external editor to work properly on Unix systems by correctly handling terminal I/O
+- Fixed external editor to show warning message when it fails to open instead of silently failing
+- Fixed find tool to properly handle no matches case without treating as error
+- Fixed find tool to wait for fd exit so error messages no longer report exit null
+- Fixed read tool to properly handle no matches case without treating as error
+- Fixed orphaned Python kernel gateway processes not being killed on process exit
+- Fixed /usage provider ordering to sort by aggregate usage (most used last)
+- Fixed /usage account dedupe to collapse identical accounts using usage metadata
+
 ## [6.9.69] - 2026-01-21
 
 ### Added
@@ -44,6 +188,7 @@
 - Fixed patch indentation normalization for fuzzy matches, tab/space diffs, and ambiguous context alignment
 
 ## [6.9.0] - 2026-01-21
+
 ### Removed
 
 - Removed Git tool and all related functionality
@@ -54,6 +199,7 @@
 - Removed @oh-my-pi/pi-git-tool dependency
 
 ## [6.8.5] - 2026-01-21
+
 ### Breaking Changes
 
 - Changed timeout parameter from seconds to milliseconds in Python tool
@@ -65,6 +211,7 @@
 - Improved streaming output handling and buffer management
 
 ## [6.8.4] - 2026-01-21
+
 ### Changed
 
 - Updated output sink to properly handle large outputs
@@ -146,6 +293,7 @@
 - Updated temporary file cleanup to use secure async removal methods
 
 ## [6.7.67] - 2026-01-19
+
 ### Added
 
 - Added normative rewrite setting to control tool call argument normalization in session history
@@ -194,7 +342,7 @@
 
 - Patch application handles repeated context blocks, preserves original indentation on fuzzy match
 - Ambiguous context matching resolves duplicates using adjacent @@ anchor positioning
-- Patch parser handles bare *** terminators, model hallucination markers, line hint ranges
+- Patch parser handles bare \*\*\* terminators, model hallucination markers, line hint ranges
 - Function context matching handles signatures with and without empty parentheses
 - Fixed session title generation to respect OMP_NO_TITLE environment variable
 - Fixed Python module discovery to use import.meta.dir for ES module compatibility
@@ -206,6 +354,7 @@
 - Fixed MCP tool path formatting to correctly display provider information
 
 ## [6.2.0] - 2026-01-19
+
 ### Changed
 
 - Improved LSP batching to coalesce formatting and diagnostics for parallel edits
@@ -243,12 +392,14 @@
 - Fixed TTSR abbreviation expansion from TTSR to Time Traveling Stream Rules
 
 ## [5.8.0] - 2026-01-19
+
 ### Changed
 
 - Updated WASM loading to use streaming for development environments with base64 fallback
 - Added scripts directory to published package files
 
 ## [5.7.68] - 2026-01-18
+
 ### Changed
 
 - Updated WASM loading to use base64-encoded WASM for better compatibility with compiled binaries
@@ -258,12 +409,14 @@
 - Fixed WASM loading issues in compiled binary builds
 
 ## [5.7.67] - 2026-01-18
+
 ### Changed
 
 - Replaced external photon-node dependency with vendored WebAssembly implementation
 - Updated image processing to use local photon library for better performance
 
 ## [5.6.70] - 2026-01-18
+
 ### Added
 
 - Added support for loading Python prelude extension modules from user and project directories
@@ -336,7 +489,7 @@
 - Enhanced Python kernel availability checking with faster validation
 - Optimized Python environment warming to avoid blocking during tool initialization
 - Reorganized settings interface into behavior, tools, display, voice, status, lsp, and exa tabs
-- Migrated environment variables from PI_ to OMP_ prefix with automatic migration
+- Migrated environment variables from PI* to OMP* prefix with automatic migration
 - Updated model selector to use TabBar component for provider navigation
 - Changed role badges to inverted style with colored backgrounds
 - Added support for /models command alias in addition to /model
@@ -369,11 +522,13 @@
 - Enhanced Python gateway environment filtering to exclude sensitive API keys and Windows system paths
 
 ## [5.5.0] - 2026-01-18
+
 ### Changed
 
 - Updated task execution guidelines to improve prompt framing and parallelization instructions
 
 ## [5.4.2] - 2026-01-16
+
 ### Changed
 
 - Updated model resolution to accept pre-serialized settings for better performance
@@ -381,11 +536,13 @@
 - Enhanced edit tool documentation with clear use cases for bash alternatives
 
 ## [5.3.0] - 2026-01-15
+
 ### Changed
 
 - Expanded bash tool guidance to explicitly list appropriate use cases including file operations, build commands, and process management
 
 ## [5.2.1] - 2026-01-14
+
 ### Fixed
 
 - Fixed stale diagnostic results by tracking diagnostic versions before file sync operations
@@ -429,6 +586,7 @@
 - Fixed session selector page up/down navigation
 
 ## [5.0.1] - 2026-01-12
+
 ### Changed
 
 - Replaced wasm-vips with Photon for more stable WASM image processing
@@ -455,6 +613,7 @@
 - Move `sharp` to optional dependencies with all platform binaries to fix arm64 runtime errors
 
 ## [4.7.0] - 2026-01-12
+
 ### Added
 
 - Add `omp config` subcommand for managing settings (`list`, `get`, `set`, `reset`, `path`)
@@ -474,6 +633,7 @@
 ## [4.6.0] - 2026-01-12
 
 ### Added
+
 - Add `/skill:name` slash commands for quick skill access (toggle via `skills.enableSkillCommands` setting)
 - Add `cwd` to SessionInfo for session list display
 - Add custom summarization instructions option in tree selector
@@ -481,10 +641,12 @@
 - Add `shutdownRequested` and `checkShutdownRequested()` for extension-initiated shutdown
 
 ### Fixed
+
 - Component `invalidate()` now properly rebuilds content on theme changes
 - Force full re-render after returning from external editor
 
 ## [4.4.8] - 2026-01-12
+
 ### Changed
 
 - Changed review finding priority format from numeric (0-3) to string labels (P0-P3) for clearer severity indication
@@ -504,6 +666,7 @@
 - Fixed frontmatter parsing to properly report source location when YAML parsing fails
 
 ## [4.4.4] - 2026-01-11
+
 ### Added
 
 - Added `todo_write` tool for creating and managing structured task lists during coding sessions
@@ -534,6 +697,7 @@
 - Fixed prompt template loading to strip leading HTML comment metadata blocks
 
 ## [4.3.2] - 2026-01-11
+
 ### Changed
 
 - Increased default bash output preview from 5 to 10 lines when collapsed
@@ -573,6 +737,7 @@
 - Fixed serialized auth storage initialization so OAuth refreshes in subagents don't crash
 
 ## [4.2.2] - 2026-01-11
+
 ### Added
 
 - Added persistent cache storage for Codex usage data that survives application restarts
@@ -593,6 +758,7 @@
 - Removed `planner` agent command template, consolidating planning functionality into the `plan` agent
 
 ## [4.2.1] - 2026-01-11
+
 ### Added
 
 - Added automatic discovery and listing of AGENTS.md files in the system prompt, providing agents with an authoritative list of project-specific instruction files without runtime searching
@@ -659,6 +825,7 @@
 - Hardened file permissions on agent database directory (700) and database file (600) to restrict access
 
 ## [4.1.0] - 2026-01-10
+
 ### Added
 
 - Added persistent prompt history with SQLite-backed storage and Ctrl+R search
@@ -668,6 +835,7 @@
 - Fixed credential blocking logic to correctly check for remaining available credentials instead of always returning true
 
 ## [4.0.1] - 2026-01-10
+
 ### Added
 
 - Added usage limit error detection to enable automatic credential switching when Codex accounts hit rate limits
@@ -771,11 +939,13 @@
 - Extension directories in `settings.json` respect `package.json` manifests
 
 ## [3.37.0] - 2026-01-10
+
 ### Changed
 
 - Improved bash command display to show relative paths for working directories within the current directory, and hide redundant `cd` prefix when working directory matches current directory
 
 ## [3.36.0] - 2026-01-10
+
 ### Added
 
 - Added `calc` tool for basic mathematical calculations with support for arithmetic operators, parentheses, and hex/binary/octal literals
@@ -794,6 +964,7 @@
 - Improved completion notification message to include session title when available
 
 ## [3.35.0] - 2026-01-09
+
 ### Added
 
 - Added retry logic with exponential backoff for auto-compaction failures
@@ -1871,15 +2042,14 @@ Total color count increased from 46 to 50. See [docs/theme.md](docs/theme.md) fo
 - **Credential storage refactored**: API keys and OAuth tokens are now stored in `~/.omp/agent/auth.json` instead of `oauth.json` and `settings.json`. Existing credentials are automatically migrated on first run. ([#296](https://github.com/badlogic/pi-mono/issues/296))
 
 - **SDK API changes** ([#296](https://github.com/badlogic/pi-mono/issues/296)):
-
-  - Added `AuthStorage` class for credential management (API keys and OAuth tokens)
-  - Added `ModelRegistry` class for model discovery and API key resolution
-  - Added `discoverAuthStorage()` and `discoverModels()` discovery functions
-  - `createAgentSession()` now accepts `authStorage` and `modelRegistry` options
-  - Removed `configureOAuthStorage()`, `defaultGetApiKey()`, `findModel()`, `discoverAvailableModels()`
-  - Removed `getApiKey` callback option (use `AuthStorage.setRuntimeApiKey()` for runtime overrides)
-  - Use `getModel()` from `@oh-my-pi/pi-ai` for built-in models, `modelRegistry.find()` for custom models + built-in models
-  - See updated [SDK documentation](docs/sdk.md) and [README](README.md)
+   - Added `AuthStorage` class for credential management (API keys and OAuth tokens)
+   - Added `ModelRegistry` class for model discovery and API key resolution
+   - Added `discoverAuthStorage()` and `discoverModels()` discovery functions
+   - `createAgentSession()` now accepts `authStorage` and `modelRegistry` options
+   - Removed `configureOAuthStorage()`, `defaultGetApiKey()`, `findModel()`, `discoverAvailableModels()`
+   - Removed `getApiKey` callback option (use `AuthStorage.setRuntimeApiKey()` for runtime overrides)
+   - Use `getModel()` from `@oh-my-pi/pi-ai` for built-in models, `modelRegistry.find()` for custom models + built-in models
+   - See updated [SDK documentation](docs/sdk.md) and [README](README.md)
 
 - **Settings changes**: Removed `apiKeys` from `settings.json`. Use `auth.json` instead. ([#296](https://github.com/badlogic/pi-mono/issues/296))
 
@@ -1910,16 +2080,15 @@ Total color count increased from 46 to 50. See [docs/theme.md](docs/theme.md) fo
 ### Added
 
 - **Compaction hook improvements**: The `before_compact` session event now includes:
-
-  - `previousSummary`: Summary from the last compaction (if any), so hooks can preserve accumulated context
-  - `messagesToKeep`: Messages that will be kept after the summary (recent turns), in addition to `messagesToSummarize`
-  - `resolveApiKey`: Function to resolve API keys for any model (checks settings, OAuth, env vars)
-  - Removed `apiKey` string in favor of `resolveApiKey` for more flexibility
+   - `previousSummary`: Summary from the last compaction (if any), so hooks can preserve accumulated context
+   - `messagesToKeep`: Messages that will be kept after the summary (recent turns), in addition to `messagesToSummarize`
+   - `resolveApiKey`: Function to resolve API keys for any model (checks settings, OAuth, env vars)
+   - Removed `apiKey` string in favor of `resolveApiKey` for more flexibility
 
 - **SessionManager API cleanup**:
-  - Renamed `loadSessionFromEntries()` to `buildSessionContext()` (builds LLM context from entries, handling compaction)
-  - Renamed `loadEntries()` to `getEntries()` (returns defensive copy of all session entries)
-  - Added `buildSessionContext()` method to SessionManager
+   - Renamed `loadSessionFromEntries()` to `buildSessionContext()` (builds LLM context from entries, handling compaction)
+   - Renamed `loadEntries()` to `getEntries()` (returns defensive copy of all session entries)
+   - Added `buildSessionContext()` method to SessionManager
 
 ## [0.27.5] - 2025-12-24
 
@@ -2138,13 +2307,13 @@ Total color count increased from 46 to 50. See [docs/theme.md](docs/theme.md) fo
 - **Custom tools now require `index.ts` entry point**: Auto-discovered custom tools must be in a subdirectory with an `index.ts` file. The old pattern `~/.omp/agent/tools/mytool.ts` must become `~/.omp/agent/tools/mytool/index.ts`. This allows multi-file tools to import helper modules. Explicit paths via `--tool` or `settings.json` still work with any `.ts` file.
 
 - **Hook `tool_result` event restructured**: The `ToolResultEvent` now exposes full tool result data instead of just text. ([#233](https://github.com/badlogic/pi-mono/pull/233))
-  - Removed: `result: string` field
-  - Added: `content: (TextContent | ImageContent)[]` - full content array
-  - Added: `details: unknown` - tool-specific details (typed per tool via discriminated union on `toolName`)
-  - `ToolResultEventResult.result` renamed to `ToolResultEventResult.text` (removed), use `content` instead
-  - Hook handlers returning `{ result: "..." }` must change to `{ content: [{ type: "text", text: "..." }] }`
-  - Built-in tool details types exported: `BashToolDetails`, `ReadToolDetails`, `GrepToolDetails`, `FindToolDetails`, `LsToolDetails`, `TruncationResult`
-  - Type guards exported for narrowing: `isBashToolResult`, `isReadToolResult`, `isEditToolResult`, `isWriteToolResult`, `isGrepToolResult`, `isFindToolResult`, `isLsToolResult`
+   - Removed: `result: string` field
+   - Added: `content: (TextContent | ImageContent)[]` - full content array
+   - Added: `details: unknown` - tool-specific details (typed per tool via discriminated union on `toolName`)
+   - `ToolResultEventResult.result` renamed to `ToolResultEventResult.text` (removed), use `content` instead
+   - Hook handlers returning `{ result: "..." }` must change to `{ content: [{ type: "text", text: "..." }] }`
+   - Built-in tool details types exported: `BashToolDetails`, `ReadToolDetails`, `GrepToolDetails`, `FindToolDetails`, `LsToolDetails`, `TruncationResult`
+   - Type guards exported for narrowing: `isBashToolResult`, `isReadToolResult`, `isEditToolResult`, `isWriteToolResult`, `isGrepToolResult`, `isFindToolResult`, `isLsToolResult`
 
 ## [0.23.4] - 2025-12-18
 
@@ -2173,13 +2342,12 @@ Total color count increased from 46 to 50. See [docs/theme.md](docs/theme.md) fo
 - Improved system prompt documentation section with clearer pointers to specific doc files for custom models, themes, skills, hooks, custom tools, and RPC.
 
 - Cleaned up documentation:
-
-  - `theme.md`: Added missing color tokens (`thinkingXhigh`, `bashMode`)
-  - `skills.md`: Rewrote with better framing and examples
-  - `hooks.md`: Fixed timeout/error handling docs, added import aliases section
-  - `custom-tools.md`: Added intro with use cases and comparison table
-  - `rpc.md`: Added missing `hook_error` event documentation
-  - `README.md`: Complete settings table, condensed philosophy section, standardized OAuth docs
+   - `theme.md`: Added missing color tokens (`thinkingXhigh`, `bashMode`)
+   - `skills.md`: Rewrote with better framing and examples
+   - `hooks.md`: Fixed timeout/error handling docs, added import aliases section
+   - `custom-tools.md`: Added intro with use cases and comparison table
+   - `rpc.md`: Added missing `hook_error` event documentation
+   - `README.md`: Complete settings table, condensed philosophy section, standardized OAuth docs
 
 - Hooks loader now supports same import aliases as custom tools (`@sinclair/typebox`, `@oh-my-pi/pi-ai`, `@oh-my-pi/pi-tui`, `@oh-my-pi/pi-coding-agent`).
 
@@ -2436,12 +2604,12 @@ _Dedicated to Peter's shoulder ([@steipete](https://twitter.com/steipete))_
 ### Changed
 
 - **Tool output truncation**: All tools now enforce consistent truncation limits with actionable notices for the LLM. ([#134](https://github.com/badlogic/pi-mono/issues/134))
-  - **Limits**: 2000 lines OR 50KB (whichever hits first), never partial lines
-  - **read**: Shows `[Showing lines X-Y of Z. Use offset=N to continue]`. If first line exceeds 50KB, suggests bash command
-  - **bash**: Tail truncation with temp file. Shows `[Showing lines X-Y of Z. Full output: /tmp/...]`
-  - **grep**: Pre-truncates match lines to 500 chars. Shows match limit and line truncation notices
-  - **find/ls**: Shows result/entry limit notices
-  - TUI displays truncation warnings in yellow at bottom of tool output (visible even when collapsed)
+   - **Limits**: 2000 lines OR 50KB (whichever hits first), never partial lines
+   - **read**: Shows `[Showing lines X-Y of Z. Use offset=N to continue]`. If first line exceeds 50KB, suggests bash command
+   - **bash**: Tail truncation with temp file. Shows `[Showing lines X-Y of Z. Full output: /tmp/...]`
+   - **grep**: Pre-truncates match lines to 500 chars. Shows match limit and line truncation notices
+   - **find/ls**: Shows result/entry limit notices
+   - TUI displays truncation warnings in yellow at bottom of tool output (visible even when collapsed)
 
 ## [0.13.1] - 2025-12-06
 
@@ -2528,13 +2696,13 @@ _Dedicated to Peter's shoulder ([@steipete](https://twitter.com/steipete))_
 ### Added
 
 - **Context Compaction**: Long sessions can now be compacted to reduce context usage while preserving recent conversation history. ([#92](https://github.com/badlogic/pi-mono/issues/92), [docs](https://github.com/badlogic/pi-mono/blob/main/packages/coding-agent/README.md#context-compaction))
-  - `/compact [instructions]`: Manually compact context with optional custom instructions for the summary
-  - `/autocompact`: Toggle automatic compaction when context exceeds threshold
-  - Compaction summarizes older messages while keeping recent messages (default 20k tokens) verbatim
-  - Auto-compaction triggers when context reaches `contextWindow - reserveTokens` (default 16k reserve)
-  - Compacted sessions show a collapsible summary in the TUI (toggle with `o` key)
-  - HTML exports include compaction summaries as collapsible sections
-  - RPC mode supports `{"type":"compact"}` command and auto-compaction (emits compaction events)
+   - `/compact [instructions]`: Manually compact context with optional custom instructions for the summary
+   - `/autocompact`: Toggle automatic compaction when context exceeds threshold
+   - Compaction summarizes older messages while keeping recent messages (default 20k tokens) verbatim
+   - Auto-compaction triggers when context reaches `contextWindow - reserveTokens` (default 16k reserve)
+   - Compacted sessions show a collapsible summary in the TUI (toggle with `o` key)
+   - HTML exports include compaction summaries as collapsible sections
+   - RPC mode supports `{"type":"compact"}` command and auto-compaction (emits compaction events)
 - **Branch Source Tracking**: Branched sessions now store `branchedFrom` in the session header, containing the path to the original session file. Useful for tracing session lineage.
 
 ## [0.12.5] - 2025-12-03
@@ -2571,11 +2739,11 @@ _Dedicated to Peter's shoulder ([@steipete](https://twitter.com/steipete))_
 ### Added
 
 - **Models**: Added support for OpenAI's new models:
-  - `gpt-4.1` (128K context)
-  - `gpt-4.1-mini` (128K context)
-  - `gpt-4.1-nano` (128K context)
-  - `o3` (200K context, reasoning model)
-  - `o4-mini` (200K context, reasoning model)
+   - `gpt-4.1` (128K context)
+   - `gpt-4.1-mini` (128K context)
+   - `gpt-4.1-nano` (128K context)
+   - `o3` (200K context, reasoning model)
+   - `o4-mini` (200K context, reasoning model)
 
 ## [0.12.0] - 2025-12-02
 

package/README.md

@@ -938,7 +938,7 @@ All tools are enabled by default. Use `--tools <list>` to restrict to a subset.
 | `lsp`         | Language Server Protocol queries (go-to-definition, references, hover) |
 | `todo_write`  | Track task progress during sessions                   |
 | `web_search`  | Search the web                                        |
-| `web_fetch`   | Fetch and render URLs                                 |
+| `fetch`       | Fetch and process URLs                                |
 | `python`      | Execute Python code in IPython kernel                 |
 | `notebook`    | Edit Jupyter notebook cells                           |
 

package/docs/hooks.md

@@ -345,10 +345,10 @@ Tool inputs:
 - `bash`: `{ command, timeout? }`
 - `read`: `{ path, offset?, limit? }`
 - `write`: `{ path, content }`
-- `edit`: `{ path, oldText, newText }`
+- `edit`: `{ path, old_text, new_text }`
 - `ls`: `{ path?, limit? }`
 - `find`: `{ pattern, path?, limit? }`
-- `grep`: `{ pattern, path?, glob?, ignoreCase?, literal?, context?, limit? }`
+- `grep`: `{ pattern, path?, glob?, ignore_case?, literal?, context?, limit? }`
 
 #### tool_result
 

package/docs/sdk.md

@@ -375,7 +375,7 @@ All tools are defined in `BUILTIN_TOOLS`:
 - `output` - Task output retrieval
 - `read` - File reading (text and images)
 - `task` - Subagent spawning
-- `web_fetch` - Web page fetching
+- `fetch` - URL fetching
 - `web_search` - Web search
 - `write` - File writing
 

package/examples/sdk/04-skills.ts

@@ -8,7 +8,7 @@
 import { createAgentSession, discoverSkills, SessionManager, type Skill } from "@oh-my-pi/pi-coding-agent";
 
 // Discover all skills from cwd/.omp/skills, ~/.omp/agent/skills, etc.
-const allSkills = discoverSkills();
+const { skills: allSkills } = await discoverSkills();
 console.log(
 	"Discovered skills:",
 	allSkills.map((s) => s.name),

package/package.json

@@ -1,6 +1,6 @@
 {
 	"name": "@oh-my-pi/pi-coding-agent",
-	"version": "6.9.69",
+	"version": "8.0.0",
 	"description": "Coding agent CLI with read, bash, edit, write tools and session management",
 	"type": "module",
 	"ompConfig": {
@@ -18,8 +18,8 @@
 			"import": "./src/index.ts"
 		},
 		"./hooks": {
-			"types": "./src/core/hooks/index.ts",
-			"import": "./src/core/hooks/index.ts"
+			"types": "./src/extensibility/hooks/index.ts",
+			"import": "./src/extensibility/hooks/index.ts"
 		}
 	},
 	"files": [
@@ -31,20 +31,21 @@
 	],
 	"scripts": {
 		"check": "tsgo -p tsconfig.check.json",
+		"format-prompts": "bun scripts/format-prompts.ts",
 		"clean": "rm -rf dist",
 		"build": "tsgo -p tsconfig.build.json && chmod +x dist/cli.js",
 		"build:binary": "bun run generate-wasm-b64 && bun build --compile ./src/cli.ts --outfile dist/omp",
 		"generate-template": "bun scripts/generate-template.ts",
 		"postinstall": "bun run generate-template || true",
 		"test": "bun test",
-		"prepublishOnly": "bun run generate-template && bun run clean && bun run build"
+		"prepublishOnly": "bun run generate-template"
 	},
 	"dependencies": {
-		"@oh-my-pi/omp-stats": "6.9.69",
-		"@oh-my-pi/pi-agent-core": "6.9.69",
-		"@oh-my-pi/pi-ai": "6.9.69",
-		"@oh-my-pi/pi-tui": "6.9.69",
-		"@oh-my-pi/pi-utils": "6.9.69",
+		"@oh-my-pi/omp-stats": "8.0.0",
+		"@oh-my-pi/pi-agent-core": "8.0.0",
+		"@oh-my-pi/pi-ai": "8.0.0",
+		"@oh-my-pi/pi-tui": "8.0.0",
+		"@oh-my-pi/pi-utils": "8.0.0",
 		"@openai/agents": "^0.3.7",
 		"@sinclair/typebox": "^0.34.46",
 		"ajv": "^8.17.1",
@@ -56,7 +57,6 @@
 		"handlebars": "^4.7.8",
 		"highlight.js": "^11.11.1",
 		"marked": "^15.0.12",
-		"minimatch": "^10.1.1",
 		"nanoid": "^5.1.6",
 		"ndjson": "^2.0.0",
 		"node-html-parser": "^6.1.13",