Assemble complete lightweight feature candidate

[KulmanD]

What changed

• combine accepted Issues Fail safely when an event tap interrupts an active gesture #21, Abort stale gesture holds without timers #23, Reset scroll smoothing state between independent scroll sequences #24, Preserve all CGEvent scroll delta representations #25, Standardize configuration defaults and remove machine-specific naming #26, and Prepare public, signed, multi-architecture distribution #27 in one final source tree
• keep rejected movement-delta Issue Use CGEvent movement deltas for gesture accumulation #22 excluded
• retain the optimized cumulative runtime below the Mach-O text-page boundary
• preserve isolated, rejected, pairwise, cumulative, and complete performance reports
• retain neutral generated config defaults and multi-architecture release tooling
• record the controlled clean-idle and maintainer-run physical validation results
• replace the README with one app-first public guide using the repository-owned icon
• add a full publication-readiness report covering history, privacy, licensing, releases, GitHub settings, and branch cleanup

Why

This is the single candidate to merge into main. The final documentation now separates measured idle performance, maintainer-observed physical behavior, source-publication decisions, and official binary-distribution requirements. It also removes the previous installation claim for architecture-specific release assets that do not yet exist as official downloads.

Validation

Build and packaging

• generated defaults check
• swift build
• swift build -c release
• native app build and ad-hoc signature verification
• helper remains 433528 bytes with a 212992-byte __TEXT segment
• shell syntax and workflow YAML checks
• GitHub Markdown rendering
• fresh lightweight sharing archive extraction and exclusion checks
• README icon and publication-readiness report included in the source archive
• no runtime, configuration, test, script, workflow, app, or packaging source changed by the launch-documentation commit

The local Command Line Tools installation does not provide the XCTest module, so the local swift test command cannot compile tests. The full GitHub Actions macOS test and dual-architecture archive jobs remain the merge gate.

Controlled clean idle

One controlled 150.086-second run at runtime commit 727060f recorded:

• zero measured CPU-time, wakeup, context-switch, and syscall growth
• two stable worker threads
• RSS flat at 10600448 bytes, about 10.1 MiB
• physical footprint flat at 3375584 bytes, about 3.22 MiB
• zero network sockets and zero normal-mode log files
• worker stopped completely after measurement

Apple's leaks utility reported seven allocations totaling 496 bytes rooted in Apple framework code. No project-owned symbol appeared in the allocation chain, and memory stayed flat. This is not represented as a universal leak-free claim.

Physical validation

On 2026-07-27, the maintainer reported that the complete manual checklist passed on one MX Master 3S connected to a MacBook Pro Mac16,8, Apple Silicon / arm64, running macOS 26.5.1 (25F80). Coverage included:

• left, right, and up gestures; one action per hold; repeated, below-threshold, interrupted, and stale-hold behavior
• no stuck modifiers, typed key characters, or Terminal escape text
• back and forward in Safari and Finder with no gesture-button leakage
• ratchet, free-spin, reversal, horizontal input, pause/resume, and bounded smoothing behavior
• unchanged MacBook trackpad scrolling and gestures
• mouse power cycle, disconnect/reconnect, sleep/wake, and worker lifecycle recovery

These are maintainer-observed results from one device and environment, not universal compatibility evidence.

Public-release audit

• current tree and every reachable Git blob scanned for common credential, key, certificate, personal-path, email, phone, config, cache, log, and crash patterns
• no credential, private key, token, personal home path, user config, or tracked build artifact found in the current tree
• no credential or revocation requirement identified in reachable history
• all 53 listed Actions run logs scanned with no sensitive match
• all 12 retained Actions validation ZIPs downloaded and scanned with no sensitive match
• no external Swift package dependency or bundled third-party asset identified
• historical config/denis-config.json contains ordinary settings only, but its filename is a public-presentation decision
• commit metadata exposes kulmand@post.bgu.ac.il; owner acceptance or a deliberate pre-publication history rewrite is required
• existing v0.2.0/v0.3.0 source assets contain the old personal filename
• existing v0.3.0 app is ad-hoc signed, arm64-only, and predates this candidate
• no license, branch protection, ruleset, Dependabot alerts, secret scanning, or code scanning is currently configured

The full evidence, MIT versus Apache-2.0 comparison, and exact blockers are in docs/PUBLICATION_READINESS.md.

Remaining owner gates

The PR can merge while the repository remains private. Do not make the repository public until the owner:

1. selects a license and exact copyright-holder name
2. decides whether the commit email and historical personal filename are acceptable
3. decides how to handle the existing v0.2.0/v0.3.0 release assets
4. configures the intended repository protection, security, community, topics, and social-preview settings
5. reviews links and public surfaces while signed out after visibility changes

Official app distribution additionally requires Developer ID signing, notarization, stapling, clean Apple Silicon and Intel tests, final checksums, and signed-out asset verification.

The deferred development-only systemEvents versus cgEvent experiment remains outside this candidate. A fixed active-input CPU/latency benchmark remains optional and unclaimed. No release was published and repository visibility was not changed.