add fullscreen touch interactions and clickable links

[kevinjosethomas]

• adds fullscreen pointer and touch support for cursor placement and per-item expansion for eng-6201.
• makes labeled markdown links clickable while retaining visible url fallbacks for eng-6126.
• preserves global detail cycling, text selection, hyperlink precedence, and inline terminal behavior.

---

Note

Medium Risk
Large surface area in fullscreen mouse dispatch, editor hit-testing, and many chat components, though behavior is covered by new tests and existing keyboard/selection paths are preserved.

Overview
Adds render-time click regions to the TUI: components expose regions after render, containers aggregate them, and the fullscreen viewport projects them through the pinned header, scrolling transcript, clipped dock, and overlays. A clean unmodified left press/release dispatches clicks (hyperlinks and selection/drag/modifier clicks keep precedence; painted overlay pixels block click-through).

Editor clicks focus the input and place the cursor from the last render layout (wrapping, scroll, prompt/header offsets, wide graphemes, atomic paste markers). Markdown labeled links stay clickable even when a visible URL fallback is shown.

In the coding-agent interactive UI, collapsed headers and summary rows (tools, bash, agent messages, branch/compaction/refinement summaries, skills, custom/injected prompts, collapsible errors, shell blocks in side questions, etc.) toggle only that item via Clickable or explicit regions; Ctrl+O still resets all conversation detail globally.

^{Reviewed by Cursor Bugbot for commit 7daebc3. Bugbot is set up for automated code reviews on this repo. Configure here.}

Note

Add fullscreen click dispatch, editor click-to-cursor, and clickable Markdown links

• Core click-region system in click-regions.ts, fullscreen.ts, and tui.ts: components expose getClickRegions after render; the fullscreen viewport projects regions through header, transcript, and dock layout and resolves unmodified left-button press/release pairs into region callbacks.
• Clickable wrapper in clickable.ts registers a click region over any child's full rendered output.
• Editor clicks in editor.ts map through saved render layout to source-line cursor positions, handling wrapping, Unicode, prompt prefixes, and atomic paste markers.
• Markdown link labels in markdown.ts now emit hyperlink click metadata in both native and fallback rendering paths.
• Interactive chat components (agent messages, bash/tool execution, branch summaries, collapsible errors, and others) use click regions to toggle expanded/collapsed state on header clicks; Ctrl+O still resets all expansions globally.
• Risk: tui.Component interface gains optional getClickRegions; out-of-tree components without it render normally but cannot dispatch clicks. Container and Box click-region caches are invalidated on child add/remove/clear and on render-cache invalidation to prevent stale regions.

Changes since #2430 opened

• Made the summary row of ExpandableEventMessage-derived components clickable to toggle expanded state [7daebc3]
• Updated test needle for locating compaction row click region [7daebc3]

^{Macroscope summarized d24970d.}

[github-actions]

Prime Agent performance — completed

PR 7daebc36 compared with main 7572880f.

Overall: 0 regressed · 0 improved · 41 no clear change.

Metric	Main	This PR	Change
Cold startup	1,749.0 ms	1,826.1 ms	≈ +77.1 ms (+4.41%)
Warm startup	1,109.7 ms	1,004.5 ms	≈ -105.2 ms (-9.48%)
Installation	14.21 s	14.84 s	≈ +0.63 s (+4.42%)
Compressed release artifacts	72.51 MB	72.54 MB	≈ +0.02 MB (+0.03%)
Installed footprint	577.65 MB	577.72 MB	≈ +0.07 MB (+0.01%)
Idle memory, summed RSS	1,272.13 MB	1,271.29 MB	≈ -0.84 MB (-0.07%)

Python runtime

Metric	Main	This PR	Change
Python kernel startup	152.4 ms	163.3 ms	≈ +10.8 ms (+7.11%)
Python cell round trip	0.577 ms	0.640 ms	≈ +0.063 ms (+10.93%)
Empty bash command	13.4 ms	14.7 ms	≈ +1.3 ms (+9.65%)
Bash git status	20.0 ms	20.6 ms	≈ +0.6 ms (+3.10%)
Bash 32 KiB output	13.7 ms	14.8 ms	≈ +1.1 ms (+7.77%)
35 cells / 9 shell calls	197.0 ms	196.0 ms	≈ -1.0 ms (-0.52%)
Python interrupt to done	1.712 ms	1.698 ms	≈ -0.014 ms (-0.81%)
Python state snapshot	27.2 ms	29.0 ms	≈ +1.8 ms (+6.76%)
Python state restore	389.6 ms	402.1 ms	≈ +12.5 ms (+3.22%)
Python idle RSS	35.67 MB	31.83 MB	≈ -3.84 MB (-10.76%)
Python RSS after pandas workload	98.09 MB	94.92 MB	≈ -3.17 MB (-3.24%)

Session transport

Metric	Main	This PR	Change
Private frame decode, 32 MiB in 8 KiB chunks	32.7 ms	36.2 ms	≈ +3.4 ms (+10.47%)

UI interactions

Metric	Main	This PR	Change
Resume large session (cold)	4,459.2 ms	4,256.4 ms	≈ -202.8 ms (-4.55%)
CPU, resume large session	7,740.0 ms	7,110.0 ms	≈ -630.0 ms (-8.14%)
Switch into large session	4,956.3 ms	4,238.3 ms	≈ -718.0 ms (-14.49%)
CPU, switch into large session	6,880.0 ms	6,280.0 ms	≈ -600.0 ms (-8.72%)
Open agents view from a session	191.9 ms	178.8 ms	≈ -13.2 ms (-6.86%)
CPU, open agents view	400.0 ms	350.0 ms	≈ -50.0 ms (-12.50%)
Full agents roster, many sessions	4.86 s	4.86 s	≈ +0.00014 s (+0.00%)
CPU, full agents roster	3.44 s	3.16 s	≈ -0.28 s (-8.14%)
Open another session from agents view	2,991.4 ms	2,766.5 ms	≈ -224.8 ms (-7.52%)
CPU, open from agents view	3,200.0 ms	2,840.0 ms	≈ -360.0 ms (-11.25%)
Reopen resident large session	601.9 ms	669.6 ms	≈ +67.7 ms (+11.24%)
CPU, reopen resident session	1,000.0 ms	1,170.0 ms	≈ +170.0 ms (+17.00%)
Open subagent session at depth 6	21,715.2 ms	21,761.9 ms	≈ +46.7 ms (+0.21%)
CPU, open subagent at depth 6	11,540.0 ms	11,180.0 ms	≈ -360.0 ms (-3.12%)
Open chain parent from agents view	4,127.9 ms	4,145.0 ms	≈ +17.2 ms (+0.42%)
CPU, open chain parent	4,120.0 ms	4,710.0 ms	≈ +590.0 ms (+14.32%)
Scheduled catalog, first request	2,341.9 ms	2,239.8 ms	≈ -102.2 ms (-4.36%)
CPU, scheduled catalog	3,910.0 ms	3,910.0 ms	≈ -4.4e-13 ms (-0.00%)
Scheduled catalog, repeated request	1,264.3 ms	1,518.5 ms	≈ +254.2 ms (+20.11%)
CPU, repeated catalog	1,730.0 ms	2,070.0 ms	≈ +340.0 ms (+19.65%)
Cold worker with three catalog scans	1,665.8 ms	1,769.9 ms	≈ +104.1 ms (+6.25%)
CPU, cold worker and scans	4,860.0 ms	4,830.0 ms	≈ -30.0 ms (-0.62%)
UI memory after interactions	2,693.92 MB	2,663.25 MB	≈ -30.67 MB (-1.14%)

Sandbox cost: ~$0.1653 — no inference calls.
Run, logs, and downloadable raw results

Methodology and samples

Main resolved at 2026-09-17T21:49:15.594096+00:00. Harness 7572880f.
Linux x64, 4 vCPU, 8 GB RAM, 20 GB disk; region us.
Image: node:24-bookworm@sha256:be23f54a88d34e8824c741b19b91064094f92c1c97b194144bfc8b50d67258e2.
Stock tools, skills, daemon, and Python bootstrap enabled; fresh homes and a fixed Git fixture.
Onboarding is dismissed; the editor starts without a selected model or submitted prompt.
Medians shown. Arrows require a 20% timing/memory change plus absolute floors and IQR.
These practical noise floors are not a statistical significance test.
Cold means stopped Prime processes; OS filesystem caches are not flushed.
No model requests or credentials. Installation excludes build/setup time.
Installer tarballs use loopback; npm/Python downloads use the network with fresh caches.
Artifact size counts release tarballs; footprint after first use includes registry packages.
MB is decimal. Summed RSS can double-count shared pages; PSS is recorded when available.
Provisioning, setup, and build durations are recorded separately in the raw results.
Kernel probes use the installed JSONL runtime, outside the TUI/TypeScript host.
Per trial: 50 Python cells, 5 calls per shell case, and one 35-cell mix (9 git status calls).
Cell/shell values are batch means; other runtime timings are single operations.
State fixture: a 10,000-row × 8-column integer DataFrame and a 10,000-integer list.
Restore runs in a fresh kernel, including pandas imports; kernel startup is excluded.
Kernel RSS covers the isolated Python process; loaded RSS follows the pandas workload.
Transport benches run node against the prepared source build, outside the installed home.
Frame decode times one 32 MiB private frame, snapshot-chunk header, pushed in
8 KiB chunks; the wire shape of multi-MB frames on the daemon-worker channels.
UI trials use a fresh fixture set: 194 top-level sessions including one ~40 MB transcript,
40 ledger fan-out children, and a 6-deep subagent chain (~46 spawn edges).
Large fixtures hold 1,999 complete triples (~5 MB JSONL); medium 119; subagents 399 each.
Interactions: cold --resume of a large session, warm /resume switch, left-arrow to agents view,
roster settle with many saved sessions, search-and-open of another large session,
reattaching to that resident session, opening the chain parent, and drilling to depth 6.
Readiness is the rendered transcript tail plus a confirmed editor echo.
CPU metrics sum utime+stime across the whole benchmark-user process tree per interaction.
UI memory sums RSS after the interactions; PTY byte counts are in the raw results.
A separate catalog fixture has 2,300 sessions, 2,298 edges, and 13 paused scheduled-job owners.
Catalog timings cover first/repeated reads and cold worker creation under three pending scans.
All expected jobs and owner metadata are checked; worker readiness excludes TUI rendering.
Costs estimate full sandbox lifetimes at configured rates, including setup and build.
Budget target: $1; not a billing cap. Performance changes are informational.
Failed or incomplete execution fails the workflow; saved artifacts remain available.
Each side stops a phase after 2 identical consecutive failures.
Skipped trials are not attempted samples. Warm startup requires a successful cold launch.

Metric	Main successful/attempted	PR successful/attempted	Main spread	PR spread
Cold startup	10/10	10/10	IQR 140.1 ms	IQR 163.1 ms
Warm startup	10/10	10/10	IQR 144.3 ms	IQR 42.7 ms
Installation	3/3	3/3	range 0.44 s	range 0.31 s
Compressed release artifacts	1/1	1/1	—	—
Installed footprint	1/1	1/1	—	—
Idle memory, summed RSS	10/10	10/10	IQR 15.40 MB	IQR 15.91 MB
Python kernel startup	10/10	10/10	IQR 20.4 ms	IQR 12.9 ms
Python cell round trip	10/10	10/10	IQR 0.086 ms	IQR 0.118 ms
Empty bash command	10/10	10/10	IQR 2.3 ms	IQR 3.4 ms
Bash git status	10/10	10/10	IQR 3.2 ms	IQR 5.1 ms
Bash 32 KiB output	10/10	10/10	IQR 3.8 ms	IQR 4.1 ms
35 cells / 9 shell calls	10/10	10/10	IQR 18.5 ms	IQR 33.7 ms
Python interrupt to done	10/10	10/10	IQR 0.115 ms	IQR 0.150 ms
Python state snapshot	10/10	10/10	IQR 2.8 ms	IQR 10.6 ms
Python state restore	10/10	10/10	IQR 42.0 ms	IQR 53.5 ms
Python idle RSS	10/10	10/10	IQR 4.22 MB	IQR 3.18 MB
Python RSS after pandas workload	10/10	10/10	IQR 3.90 MB	IQR 3.77 MB
Private frame decode, 32 MiB in 8 KiB chunks	10/10	10/10	IQR 3.8 ms	IQR 3.5 ms
Resume large session (cold)	3/3	3/3	range 1,178.1 ms	range 919.9 ms
CPU, resume large session	3/3	3/3	range 1,760.0 ms	range 1,670.0 ms
Switch into large session	3/3	3/3	range 1,151.5 ms	range 217.6 ms
CPU, switch into large session	3/3	3/3	range 980.0 ms	range 870.0 ms
Open agents view from a session	3/3	3/3	range 16.8 ms	range 16.6 ms
CPU, open agents view	3/3	3/3	range 420.0 ms	range 180.0 ms
Full agents roster, many sessions	3/3	3/3	range 0.41 s	range 0.41 s
CPU, full agents roster	3/3	3/3	range 0.67 s	range 0.72 s
Open another session from agents view	3/3	3/3	range 381.2 ms	range 204.6 ms
CPU, open from agents view	3/3	3/3	range 790.0 ms	range 360.0 ms
Reopen resident large session	3/3	3/3	range 131.9 ms	range 99.4 ms
CPU, reopen resident session	3/3	3/3	range 400.0 ms	range 340.0 ms
Open subagent session at depth 6	3/3	3/3	range 1,469.8 ms	range 301.0 ms
CPU, open subagent at depth 6	3/3	3/3	range 2,600.0 ms	range 430.0 ms
Open chain parent from agents view	3/3	3/3	range 359.1 ms	range 179.2 ms
CPU, open chain parent	3/3	3/3	range 670.0 ms	range 290.0 ms
Scheduled catalog, first request	3/3	3/3	range 535.9 ms	range 169.0 ms
CPU, scheduled catalog	3/3	3/3	range 960.0 ms	range 550.0 ms
Scheduled catalog, repeated request	3/3	3/3	range 316.8 ms	range 170.9 ms
CPU, repeated catalog	3/3	3/3	range 450.0 ms	range 260.0 ms
Cold worker with three catalog scans	3/3	3/3	range 339.3 ms	range 360.5 ms
CPU, cold worker and scans	3/3	3/3	range 940.0 ms	range 940.0 ms
UI memory after interactions	3/3	3/3	range 73.74 MB	range 80.55 MB

[cursor]

Stale Bugbot comment from a previous run.

[cursor]

Cursor Bugbot has reviewed your changes and found 1 potential issue.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

^{Reviewed by Cursor Bugbot for commit d24970d. Configure here.}