Research Watchlist
This page is the contributor-facing index of external projects, papers, product docs, and implementation references that jackin' keeps revisiting when designing roadmap items. It is not an endorsement list and it is not a dependency list. It is a reminder to verify prior art before inventing a local shape, and to record the snapshot date when a roadmap item borrows an idea.
Use this page before implementation work that touches agent orchestration, runtime status, containment, credentials, desktop surfaces, or automation. Re-check the relevant upstream source, docs, license, and recent issues; then summarize the parts that influenced the jackin' design in the specific roadmap item or PR. Do not copy incompatible source, and do not assume a project snapshot from an older roadmap page is still current.
Agent Runtime Status And Terminal Observability
| Reference | Watch for | Where it informs jackin' |
|---|---|---|
| Herdr / docs | Foreground process detection, current-screen state detection, socket status reports, attention roll-up, done until viewed, and issue tracker edge cases around stale reports. | Agent runtime status authority and jackin' Capsule. |
| multicode | Codex app-server status, OpenCode SSE/session status, descendant session roll-up, cooperative state files, GitHub/task queue surfaces, and Apple container experiments. | Agent runtime status, Agent workflow orchestration, Autonomous task queue, GitHub link tracking, and Agent Orchestrator Research Program. |
| CCManager | xterm/headless visible-screen classifiers, Claude idle debounce, prompt-box-region parsing, approval-state handling, fixture strategy, and fast polling trade-offs. | Agent runtime status screen-detector tests and anti-flicker behavior. |
| Agent Session Manager | tmux capture-pane parsing, spinner detection, followed-window roll-up, waiting-over-busy priority, and wide-character/line-cleanup details. | Status roll-up and regression fixtures for terminal-text classifiers. |
| WezTerm Agent Deck | Foreground process lookup in terminal panes, child process fallback, pane-title fallback, recent-line priority, notification TTLs, and rate limiting. | Process identity fallback, cache TTLs, and attention notification policy. |
| ccmux | needs_input transitions, keep-awake/sleep-inhibit behavior, prompt counts, real tmux/PTY e2e tests, and the limits of quiet-pane classifiers for non-Claude agents. | Agent attention prompts, idle cleanup, and status-to-operational-effect mapping. |
| TUICommander | Foreground process polling, status-line markers, session-binding environment variables, shell wrappers, terminal activity state machines, subtask counters, and completion deferral. | Parent/child agent roll-up, reporter launch environment, and completion safety. |
| Codemux agent status and tmux-agent-status | Hook-driven lifecycle events, local status files, sidebar summaries, done-until-viewed UX, and hook-over-process authority. | Container-local hook reporters and acknowledgement semantics. |
| agent-deck, claudeye, and Codeman | Transition notifications, hook freshness windows, scratch config ownership, done sentinels, health scores, circuit breakers, AI idle-check fallbacks, and terminal classifier variants. | Adjacent ideas for status evidence, failure handling, stuck diagnostics, and notification throttling. |
| Zylos cursor-readiness research | Cursor-position readiness probes, output-stability windows, and regex failure modes. | Input-ready and stuck-detection fallback design. |
| VS Code terminal shell integration and Windows Terminal shell integration | Prompt/command OSC markers, command-finished events, cwd reporting, and shell integration fallback semantics. | Shell prompt/execution boundaries inside Capsule PTYs. |
| Warp | Agent-in-terminal product shape, interactive command handoff, and terminal-first UX direction. | Product boundary for visible foreground agents instead of hidden workers. |
| Zellij and Ghostty | Multiplexer UX, pane rendering, and screen-state representation. | Capsule multiplexer design and screen detector substrate. |
Agent Workflow And Fleet Orchestration
| Reference | Watch for | Where it informs jackin' |
|---|---|---|
| Orca and the Orca research item | Agent development environment concepts, workspace/task handoff, and fleet orchestration UX. | Agent Orchestrator Research Program and fleet-phase roadmap items. |
| Conductor, Contrabass, and agtx | Multi-agent task routing, workspace grouping, and terminal/session control surfaces. | Agent workflow orchestration and console agent session control. |
| Forge MCP and Forge Terminal MCP | MCP-driven agent operations, terminal mediation, and command/action bridges. | Custom operator tools and task-source abstraction. |
| MCO/Hive, Emdash, Vibe Kanban, Sandcastle, worktrunk, workmux, par, Optio, Sculptor, Handler.dev, Helmor, OpenClaw Code Agent, RuFlo, Ruah/Bernstein, Switchboard, and GitHub Agentic Workflows | Queue models, PR/task lifecycle, delegation semantics, agent fleets, and project-board-style operator surfaces. | Agent workflow orchestration and Autonomous task queue. Re-resolve project URLs from the roadmap item before use because this group changes quickly. |
| Agent Client Protocol | Runtime-neutral agent/session protocol ideas and ecosystem convergence. | Future cross-runtime control contracts and custom agent integration. |
Isolation, Security, And Runtime Backends
| Reference | Watch for | Where it informs jackin' |
|---|---|---|
| Hazmat | Dedicated user isolation, seatbelt/firewall tiers, integration capability rules, Docker boundary analysis, harnesses, and threat matrix structure. | Agent Orchestrator Research Program, Container credential exposure, and sandboxing roadmap items. |
| Docker Sandboxes | Branch-mode worktrees, signed commits, hypervisor isolation, credential boundaries, and sandbox lifecycle. | Selectable sandbox backends, Per-mount isolation, and Docker runtime hardening work. |
| OrbStack isolated machines and related issue threads | Fast local VM/container ergonomics, isolated Docker engines, and macOS backend constraints. | OrbStack isolated machine backend and selectable sandbox backends. |
| smolvm, smol-machines SDK, and libkrun | Lightweight VM backend feasibility, filesystem/network constraints, and Rust integration risk. | smolvm backend research and future microVM backend design. |
| Zerobox and OpenAI Codex CLI sandboxing behavior | Process-level sandbox controls, command mediation, and host/guest trust boundaries. | Process-level sandboxing and command-execution safety design. |
| Claude Code devcontainer guidance and Docker runtime/security docs | Baseline devcontainer patterns, Docker hardening knobs, network/credential limits, and explicit non-goals. | Devcontainer parity, network egress policy, and Docker runtime hardening. |
Desktop, Accounts, And Operator Surfaces
| Reference | Watch for | Where it informs jackin' |
|---|---|---|
| CodexBar, OpenUsage, RepoBar, and GitHub pull inbox tools | Menu-bar account surfaces, usage counters, repo/PR notifications, and lightweight background status. | jackin' Desktop Agent Hub and token/cost telemetry. |
| Apple Human Interface Guidelines, menu bar guidance, and Liquid Glass guidance | Native macOS menu bar behavior, status-item constraints, and visual-system expectations. | Desktop Agent Hub UX constraints. |
| Microsoft APM | Native authentication/account management patterns for agent platforms. | Native APM role support and authentication UX. |
Supporting Infrastructure And Documentation Tools
| Reference | Watch for | Where it informs jackin' |
|---|---|---|
| 1Password CLI, Doppler, and Vault Agent | Secret injection patterns, env templating, lease/refresh semantics, and credential sidecar trade-offs. | Container credential exposure and credential-source pattern work. |
| rumdl | Fast Markdown linting, rule coverage, and docs CI ergonomics. | Docs markdown linting. |
| zot | Local registry/cache behavior and OCI distribution mechanics. | Workspace registry cache. |
Git interpret-trailers and GitHub co-authored-by docs | Trailer parsing, attribution conventions, and generated commit metadata. | Public attribution and project growth. |
Maintenance Rules
- Add a project here when at least two roadmap items keep referring to it, or when one roadmap item depends on repeated verification of that project's design.
- Keep implementation details in the specific roadmap item. This page should say what to watch and where it matters, not duplicate every research note.
- Record snapshot dates in the roadmap item or PR description when a design decision depends on a specific upstream behavior.
- Prefer primary sources: repository source, official docs, release notes, and issue discussions. Blog posts are useful context but should not be the only evidence for an implementation detail.
- Re-check licenses before using source-level ideas. AGPL and proprietary sources can inform product shape, but code for jackin' must be written independently.