SShortSingh.
TopTechnologyAI & MLStartupsProgrammingScienceBusinessFinanceCrypto & Web3World
Back to feed
Programming

12-Year-Old Developer's Security Tool Blocks Its Own Creator for Two Days

0
·5 views

A 12-year-old developer built KAVLR, a security tool meant to act as a leash for AI agents, and tested it on a $150 Android phone using Termux. Its fail-closed design blocked commands whenever the backend server was unreachable, and attempts to host the backend on Vercel and Railway failed before Supabase Edge Functions worked. The tool then locked the developer out of their own terminal and files, creating what they called a zombie process that resisted deletion and corrupted their code. After tracing the problem with grep and rebuilding the affected files and symlinks, the developer restored normal function, and the hook correctly blocked a test command for a .env file. The developer shared the account as a lesson in persistence for others facing similar debugging battles.

Read the full story at DEV Community

This is an AI-generated summary. ShortSingh links to the original source for the complete article.

Discussion (0)

Log in to join the discussion and vote.

Log in

Related stories

0
ProgrammingDEV Community ·

Browser-Based File Converter Keeps Images Local, Avoiding Server Uploads

The article describes a free image converter that processes files entirely in the browser rather than uploading them to a server. The author argues that typical online converters expose sensitive files to infrastructure users cannot verify, and offers a simple network-inspection method to confirm no file is transmitted. Conversion relies on decoding with createImageBitmap, drawing to a canvas, and encoding through WebAssembly codecs running in a Web Worker. The piece also covers practical challenges, including EXIF rotation, worker timeouts, HEIC decoding under strict Content-Security-Policy, and binary-searching JPEG quality to meet target file sizes.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Causal Read-Your-Own-Writes in Java Using Postgres LSN Tokens Instead of Primary Routing

The article argues that sending reads to the primary database right after a write is a costly shortcut that limits write capacity at scale. It recommends capturing the Postgres log sequence number (LSN) during the mutation transaction and returning it to the client, for example as an HTTP header. Subsequent reads carry this token, and a routing layer sends them to a replica only if the replica has replayed past that position. A fallback to the primary applies when replication lag exceeds a set threshold, such as 25ms. The author presents this as a lighter alternative to distributed locks or consensus protocols when read-your-own-writes consistency is the only requirement.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Claude Code skills can be silently dropped from context when their listing budget overflows

Claude Code includes every installed skill's name and description in the model's context on each turn, whether or not it is used. The listing is capped at 1% of the context window, and when it overflows, descriptions of the least-used skills are removed without any UI warning, only a debug log entry. Skills left without descriptions can no longer be matched to user requests and effectively stop being selected. The author built skill-tax-cli, a dependency-free Python tool that measures each skill's token cost, replays the overflow rule, and exits non-zero when skills would lose their descriptions, making it usable in CI. The tool also measures rules, memory files, and MCP configs, and token counts are rough estimates based on characters divided by four.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Developer Adds Job Posting Health Check to Resume Tool to Flag Suspicious Listings

A developer behind the ResumeIQ AI tool argues that job seekers should assess whether a posting is worth applying to before optimizing their resume. The author built a feature called Job Posting Health that scans a pasted job description for warning signs such as registration fees, pressure to act quickly, and required equipment purchases from a specified seller. In testing with a sample posting containing several such signals, the feature flagged them for closer review. The developer stresses the tool cannot definitively determine whether a job is legitimate, only prompt users to investigate further. A companion feature, Resume Freshness, checks whether a resume shows recent evidence of the skills a role requires.

0 comments Read more at DEV Community