惯性聚合 高效追踪和阅读你感兴趣的博客、新闻、科技资讯
阅读原文 在惯性聚合中打开

推荐订阅源

奇客Solidot–传递最新科技情报
奇客Solidot–传递最新科技情报
Jina AI
Jina AI
博客园 - Franky
Apple Machine Learning Research
Apple Machine Learning Research
酷 壳 – CoolShell
酷 壳 – CoolShell
阮一峰的网络日志
阮一峰的网络日志
量子位
雷峰网
雷峰网
宝玉的分享
宝玉的分享
V
Visual Studio Blog
博客园_首页
小众软件
小众软件
The Cloudflare Blog
让小产品的独立变现更简单 - ezindie.com
让小产品的独立变现更简单 - ezindie.com
大猫的无限游戏
大猫的无限游戏
博客园 - 聂微东
S
SegmentFault 最新的问题
博客园 - 【当耐特】
OSCHINA 社区最新新闻
OSCHINA 社区最新新闻
博客园 - 叶小钗
月光博客
月光博客
博客园 - 三生石上(FineUI控件)
人人都是产品经理
人人都是产品经理
WordPress大学
WordPress大学

Hacker News: Show HN

PurrrrrFocus: Pomodoro Timer App - App Store Workflow Engine — Multi-Step Orchestration for Bun RapidPhoto: Pro Photo Editor App - App Store GitHub - DheerG/swarms: Achieve extraordinary results with claude code across a variety of tasks SPICE simulation → oscilloscope → verification with Claude Code — Lucas Gerads Show HN: VCoding – A 5 MB native Windows IDE with no dynamic dependencies Show HN: LLMs don't hallucinate because they're bad at math, it's the format GitHub - Agent-FM/agentfm-core: AgentFM is a peer-to-peer network that turns everyday computers into a decentralized AI supercomputer. AgentFM lets you run massive AI workloads directly across a global mesh of idle CPUs and GPUs. Show HN: Tracking Top US Science Olympiad Alumni over Last 25 Years GitHub - Potarix/agent-hub: One place to talk to all your agents Show HN: Runtime security for AI agents(injection,tool abuse, data exfiltration) GitHub - dubeyKartikay/lazyspotify: Terminal Spotify client for macOS and Linux GitHub - the-banana-tool/king-louie: Easy to use GUI Personal AI Assistant. Win/Linux/Mac. Show HN I made my vacation rental bookable by AI agents–no Airbnb, 0% commission GitHub - basteez/jsf-autoreload: maven plugin to enable hot reload on jsf projects uvm32/hosts/host-gdbstub at main · ringtailsoftware/uvm32 GitHub - labsai/EDDI: Config-driven engine that turns JSON into production-grade AI agents. Multi-agent orchestration, 12+ LLM providers, MCP/A2A protocols, RAG, persistent memory, and enterprise compliance (EU AI Act, GDPR, HIPAA). Built on Quarkus. GitHub - glitchnsec/fortyone-oss: AI Executive Assistant Platform Quickstart | Alien GitHub - muxshed/shed: One stream in, or many. Every destination, simultaneously. No cloud middleman, no per-channel fees, no limits. GitHub - ocrbase-hq/ocrbase: 📄 PDF/IMG ->.MD/JSON Document OCR API for PaddleOCR and GLMOCR. Self-hostable. GitHub - impactjo/home-memory: MCP server that lets your AI assistant remember everything about your home. GitHub - Sets88/dbcls: DbCls is a powerful terminal database client that supports various databases GitHub - neptun2000/heor-agent-mcp GitHub - SeanFDZ/macmind: Single-layer transformer in HyperTalk for the classic Macintosh RollQuation: Math Puzzles - Apps on Google Play GitHub - dropbox/witchcraft Show HN: Agent-cache – Multi-tier LLM/tool/session caching for Valkey and Redis GitHub - opentalon/opentalon: OpenTalon is an open-source platform built from the ground up in Go as a robust alternative to OpenClaw LinkedIn™ 职位抓取工具 - Chrome 应用商店
GitHub - jftuga/mac-screen-search: macOS CLI tool that ca...
jftuga · 2026-04-29 · via Hacker News: Show HN

Code Base: AI Vibes

A macOS CLI tool that captures a screenshot of the entire screen, performs OCR to find all instances of a search term, draws colored rectangles around each match, and opens the annotated image in Preview.

It can also process existing image files in batch using a file glob pattern, making it useful for redacting or blurring sensitive text across many screenshots at once.

screenshot mode redact mode
screenshot mode redact mode

AI Disclaimer

This project was vibe-coded with Claude Opus 4.6. While it has been tested on the author's system, it interacts directly with macOS screen capture, image processing, and file I/O, the author makes no guarantees and assumes no responsibility for unintended behavior, missed redactions, or any other issues arising from its use. Use at your own risk and always verify redactions before sharing sensitive screenshots.

Requirements

  • macOS (uses ScreenCaptureKit, Vision, and CoreGraphics)
  • Screen Recording permission (for screenshot mode)
  • Swift compiler (only needed if compiling from source; otherwise install via Homebrew or download a release)

Installation

brew tap jftuga/tap
brew install jftuga/tap/mac-screen-search

To upgrade to a newer version:

brew update
brew upgrade mac-screen-search

Manual download

Download the macOS arm64 binary from the releases page:

tar -xJf mac-screen-search_v*.tar.xz
cd mac-screen-search_v*/
sudo cp mac-screen-search /usr/local/bin/

Building

swiftc -o mac-screen-search mac-screen-search.swift

Usage

mac-screen-search [-r] [-b <pct>] [-e] [-d <dist>] [-c <color>] [-v]
                  [-n] [-o <path>] [-m <n|all>] [-M] [-l] [-t <secs>] [-w]
                  [-D <delim>] <search-term> [-f <glob>]

Options

Flag Description
-r Redact (fill with solid color) matched regions instead of outlining them
-b <percent> Blur matched regions with Gaussian blur (1-100); mutually exclusive with -r
-e Enhanced OCR (preprocess image + check multiple candidates)
-d <dist> Fuzzy match using Levenshtein distance threshold
-c <color> Rectangle color name (default: red). Available: black, blue, cyan, gray, green, magenta, orange, pink, purple, red, white, yellow
-f <glob> Process image files matching the glob pattern instead of capturing the screen
-n Do not open the result in Preview (screenshot mode only)
-o <path> Output directory or file path for the screenshot PNG (screenshot mode only)
-m <n|all> Capture monitor n (1-based) or all monitors (screenshot mode only)
-M List connected monitors and exit
-l List matches (text and coordinates) without annotating; incompatible with -r/-b
-t <seconds> Capture delay in seconds (default: 2; use 0 for immediate). Screenshot mode only
-w Whole-word matching (require word boundaries on both sides of the match)
-D <delim> Delimiter for multiple search terms (default: |). The search-term argument is split on this delimiter; each piece is searched independently and all matches are merged
-h Print help and exit
-v Print version and exit

Screenshot mode (default)

Capture a screenshot, find all occurrences of the search term, outline them in red (default), and open the result in Preview:

mac-screen-search "password"

There is a 2-second delay before capture so you can arrange your screen. Use -t to change it:

mac-screen-search -t 5 "password"    # 5-second delay
mac-screen-search -t 0 "password"    # immediate capture

Custom color

Use blue rectangles instead of red:

mac-screen-search -c blue "password"

Redact mode

Fill matched regions with a solid color to obscure the text:

mac-screen-search -r "my_password"
mac-screen-search -r -c black "my_password"

Blur mode

Apply a Gaussian blur to matched regions instead of filling or outlining them. The percentage controls blur intensity (1 = subtle, 100 = fully obscured):

mac-screen-search -b 50 "my_password"
mac-screen-search -b 80 "api-key" -f '*.png'

The -b and -r flags are mutually exclusive.

No-open mode

Save the annotated screenshot without opening it in Preview. Useful for scripting and automation:

mac-screen-search -n "password"

Output path

Control where the screenshot PNG is saved instead of the current working directory:

mac-screen-search -o ~/Screenshots "password"          # save to directory
mac-screen-search -o ~/Screenshots/result.png "password"  # save to specific file

When used with -m all, -o must point to a directory.

Monitor selection

Capture a specific monitor by its 1-based index, or all connected monitors:

mac-screen-search -m 1 "password"      # primary monitor
mac-screen-search -m 2 "password"      # second monitor
mac-screen-search -m all "password"    # all monitors (one file per display)

Use -M to list connected monitors and their resolutions:

mac-screen-search -M

Example output:

Monitor 1: 1512x982  (3024x1964 px) [displayID: 1]
Monitor 2: 2560x1440 (5120x2880 px) [displayID: 2]

List mode

Print match text and pixel coordinates without annotating or saving an image. Output is tab-separated with a header row:

mac-screen-search -t 0 -l "test"
Searching for: "test"
Screenshot captured (2940x1912)
Found 3 matches
text    x       y       width   height
test    1766    183     92      42
test    1060    1544    136     35
test    331     1591    143     38

In file glob mode, a file column is prepended:

mac-screen-search -l "test" -f '*.png'
Processing 2 files matching "*.png" for "test"
file    text    x       y       width   height
/Users/john/Screenshots/a.png   test    1089    273     149     34
/Users/john/Screenshots/b.png   test    323     315     151     41

The -l flag is incompatible with -r and -b.

Whole-word matching

Only match when the search term has word boundaries on both sides. Reduces false positives for short search terms:

mac-screen-search -w "is"    # matches "is" but not "this" or "island"

Also works with fuzzy matching (-d).

Multi-term search

Search for multiple terms at once by separating them with | (the default delimiter). All terms are matched against the same OCR output and results are merged:

mac-screen-search "password|secret|api-key"

Use -D to change the delimiter if your search terms contain the | character:

mac-screen-search -D "," "password,secret,api-key"

Since the shell interprets | as a pipe, the search-term argument must be quoted.

File glob mode

Process existing image files instead of capturing a screenshot. The glob is expanded by the tool itself (not the shell), so quote the pattern:

mac-screen-search "secret" -f '*.png'

Combine with -r or -b to batch-redact or blur sensitive text across many files:

mac-screen-search -r "api-key" -f '~/Screenshots/*.png'
mac-screen-search -b 60 "api-key" -f '~/Screenshots/*.png'

In file mode:

  • Files are overwritten in-place with the annotated/redacted version
  • The original file modification time (mtime) is preserved
  • Supported image formats include PNG, JPEG, TIFF, BMP, GIF, and others supported by CGImageSource
  • A per-file summary is printed; no files are opened in Preview

Enhanced OCR mode

Use -e to improve OCR accuracy on degraded images (e.g., screenshots taken over Zoom, transparent terminal backgrounds):

mac-screen-search -e "password"
  • Composites the image onto a white background, removing transparency artifacts that confuse the OCR engine
  • Boosts contrast (1.4x) and sharpens edges to make text crisper
  • Checks the top 5 OCR candidates per text region instead of only the top 1, catching cases where the correct reading is a lower-ranked alternative

Fuzzy matching mode

Use -d <dist> to match strings within a Levenshtein (edit) distance of the search term. This catches OCR misreads where characters are substituted (e.g., Z read as 2, g as q):

mac-screen-search -d 3 "rBZrS6gq7NsD"
  • Slides a window across each recognized text line and accepts any substring within the given edit distance
  • A distance of 1 allows one character substitution, insertion, or deletion; higher values are more permissive
  • Short search terms with high distance values may produce false positives; keep the ratio reasonable

Combining flags

All flags compose freely. For example, to batch-redact a string across many screenshots with enhanced OCR and fuzzy matching:

mac-screen-search -r -e -d 3 -c black "api-key" -f '~/Screenshots/*.png'

Capture immediately, whole-word match, save to a specific directory without opening Preview:

mac-screen-search -t 0 -n -w -o ~/Screenshots "password"

Example output (file mode)

Processing 4 files matching "*.png" for "api-key"
/Users/john/Screenshots/dashboard.png: 2 matches
/Users/john/Screenshots/settings.png: 1 match

2 files updated, 0 errors

How it works

  1. Capture -- Takes a Retina-resolution screenshot via ScreenCaptureKit (or loads image files in -f mode). Supports selecting a specific monitor (-m) and configurable delay (-t)
  2. OCR -- Runs Apple Vision's VNRecognizeTextRequest with accurate recognition and language correction. With -e, the image is preprocessed first and multiple candidates are evaluated
  3. Search -- Finds all case-insensitive occurrences of each search term (split by -D delimiter, default |), mapping each to pixel-coordinate bounding boxes. With -d, uses Levenshtein distance for fuzzy matching. With -w, requires word boundaries
  4. Annotate -- Draws colored outline rectangles (or solid fill with -r, or Gaussian blur with -b) around each match, using the color specified by -c (default: red). Skipped in list mode (-l)
  5. Output -- Saves the result as a timestamped PNG and opens it in Preview (screenshot mode), or overwrites files in-place preserving mtime (file mode). Use -o to set the output path and -n to skip Preview

Personal Project Disclosure

This program is my own original idea, conceived and developed entirely:

  • On my own personal time, outside of work hours
  • For my own personal benefit and use
  • On my personally owned equipment
  • Without using any employer resources, proprietary information, or trade secrets
  • Without any connection to my employer's business, products, or services
  • Independent of any duties or responsibilities of my employment

This project does not relate to my employer's actual or demonstrably anticipated research, development, or business activities. No confidential or proprietary information from any employer was used in its creation.