Find your next challenge.

The problem

Accept images from native inputs or captured emulator, browser, and application environments, then use vision-capable models for iterative UI and rendering verification and correction.

Possibly addressed by:Vision Toolkit · Llm Vision · Computer Use Vision

The problem

Provide a blocking wait or join primitive for parallel and continuable subagents, with optional timeouts, so parent agents can await results without premature responses or repeated polling.

Possibly addressed by:Wait Subagent · Wait Guard · Background Agents

The problem

Allow asynchronously spawned or forked subagents to inherit the parent’s cached conversation context, including the current request and relevant tool state, while building their own context and safely handling inherited unfinished executions. The main agent should be able to monitor and interact with them dynamically.

Possibly addressed by:Subagent Model · Fork Relink · Agent Chat

The problem

Detect when a multi-step task or announced action remains unexecuted and keep the turn open to drive the remaining steps automatically. This prevents premature completion without requiring another manual turn.

Possibly addressed by:Loop Continue · Wait Guard · Auto Stop

The problem

Provide centralized Web controls to browse, delete, and restore archived sessions, without requiring file editing or command-line operations.

Possibly addressed by:Session Admin · Archived Sessions · Session Delete

The problem

Add configurable limits, summarization, visibility, and cancellation for tool outputs, background search requests, and sub-agent token/step usage, with warnings and circuit breakers to prevent runaway context growth and wasted spend. Also isolate sub-agent temporary directories and improve document handling and redirect support.

Possibly addressed by:Compact Agents · Tool Budget · Tool Call Limit

The problem

Expose a configurable maxRequestImages limit for llm-pi-ai provider profiles, applied alongside the existing byte budget to retain only the newest N images. This prevents requests from exceeding backend media-count limits when image sizes vary.

Possibly addressed by:Image Count Guard

The problem

Allow profiles or workspaces to define permitted filesystem roots, restricting reads and other agent or plugin file operations to those roots unless explicitly authorized. Preserve unrestricted behavior when no roots are configured and clearly report paths that are unavailable.

Possibly addressed by:Write Protect · Workspace Envrc · Multi Workspace

The problem

Retain important user and delegation instructions through context compaction, long contexts, and new sessions, while automatically restoring preferences such as language and output style.

Possibly addressed by:Agent Instructions As Prompt · Agent Instructions Editor · Agent Instructions

The problem

Add a language preference that controls model responses, displayed reasoning, and intermediate content, allowing users to select or default to Chinese without mixed-language output.

Possibly addressed by:Reply Language · Thinking Language · Language Control

The problem

Allow ACP, workspace/session APIs, and headless CLI session creation or resumption to select and mount an agent preset, including presets from `.dsh/.agent-presets/`, so the session receives its configured tools, prompts, and skills.

Possibly addressed by:Preset · Preset Agent Manager · Workspace Presets

The problem

Preserve complete conversation, memory, tools, prompts, workspace, and unfinished-work state when switching channels, agents, models, or workflows. Use provider-independent transcripts and centralized project configuration so sessions can move without manual copying or silent rebuilding.

Possibly addressed by:Background Agents · Agent Mode Switcher · Subagent Model Picker

The problem

Import historical or externally exported agent conversations, preserving relevant messages, tool calls, reasoning, and results so users can recover context and continue working.

Possibly addressed by:Chat Import · Session Import Codex · Session Import

The problem

Ship native Windows ARM64 desktop builds with installers and update artifacts, optionally including preview or nightly releases, so users can avoid x64 emulation.

The problem

Support long-running tasks that repeatedly inspect, improve, and verify their outputs until they satisfy explicit quality requirements. Use checkpoints and quality gates to detect premature completion and trigger further iteration.

Possibly addressed by:Apollo · Agent Dag Workflow · Subagent Watchdog

The problem

Provide a centralized, customizable shortcut system where plugins can register bindings and callbacks, while users can view, edit, and resolve conflicts among shortcuts. Expose stable public APIs for shortcut integration.

Possibly addressed by:Keyboard Manager · Hotkeys Platform · Shortcuts

The problem

Make web search follow the selected conversation model or session LLM route, including its provider and credentials, with support for custom routing headers. This avoids sending searches to an unrelated global endpoint and enables model-specific search backends.

Possibly addressed by:Web Search Custom · Web Search Session Follow · Web Search

The problem

Detect Windows workspace and sandbox ACL problems such as missing WRITE_OWNER, distinguish relevant DACL/SACL failures, and provide actionable guidance or repair before restricted operations fail. Document prerequisites and clearly identify any reduced-isolation fallback.

Possibly addressed by:Acl Sandbox Patch · Workspace Acl Allow · Windows Workspace Guard

The problem

Validate and, where possible, correct malformed MCP and other tool calls so open models can use tools reliably. Return actionable errors that identify the rejected tool and received top-level argument keys, including for nested calls, to prevent blind retries.

Possibly addressed by:Tool Call Guard · Toolcall Compat · Tool Error Hints

The problem

Let users enter editing or rejection feedback directly while viewing a plan, then request a revised plan without switching to chat.

Possibly addressed by:Plan Plus · Plugin Better Plan · Improved Inline Edit