nexika

by Loai Elattar

Nexika amin: a repository maintainer that never merges for you. Issue triage with approval, issues worked in isolated worktrees into pull requests, change notes enforced in CI, and releases (versions, CHANGELOG, tags, GitHub Releases) prepared and published step by step.

by Loai Elattar

Nexika barq: many file and project operations in one call. Batched reads, symbol and outline reads, a seen-before cache, short test/build output, git status with next steps, secret masking and savings stats.

by Loai Elattar

Nexika bayan: Claude writes like a clear, friendly person, in English and Arabic. Plain explanations for readers with no coding experience, automatic clean-up of machine habits (hidden characters, AI signature lines, filler phrases), and a line-by-line style check.

by Loai Elattar

Nexika hafiz: Claude remembers your work. Decisions, tasks, problems, files and links captured as you work (no AI calls), a short start card, a snapshot restored after compaction, automatic handoff notes, Arabic and English search, secrets replaced before saving, and a detailed session summary naming the issue and branch on request.

by Loai Elattar

Nexika haris: guards your machine and your accounts from harmful agent actions. Every tool call is read by a real shell parser (wrappers, pipes, substitutions and heredocs included) and judged by its action and target: safe reads and project runs are approved, risky actions are asked about, dangerous ones are refused with a plain reason. Secrets, persistence spots and haris itself are protected, data leaving the machine is checked for secrets, prompt injection in tool output is flagged, and approvals count only when you type them.

by Loai Elattar

Nexika itqan: quality workflow without the friction - plan, failing tests first, implement, verify, specialist review; a risk-based guard that only stops dangerous actions; stack checklists loaded only for the stacks a project uses.

by Loai Elattar

Nexika lawha: frontend you can see. Checks every page at six screen widths, light and dark, LTR and RTL, with reduced motion: sideways scrolling, clipped and overlapping text, tap targets, accessibility (axe), layout shift, motion, physical CSS that breaks Arabic, plus an artistic eye that measures alignment, spacing rhythm, type scale and colour. Indexes the project's design system (Tailwind tokens, shadcn, components with props, TanStack routes). Built for React, TanStack, Tailwind and shadcn/ui. English and Arabic.

by Loai Elattar

Nexika manar: be found by search engines and AI assistants. Deterministic SEO and AI-visibility audits, fixes written into your code (Next.js, Astro, ASP.NET Core, static sites) through reviewed pull requests, and real measurement of whether ChatGPT, Gemini, Perplexity and Claude mention and cite you.

by Loai Elattar

Nexika mizan: keeps your Claude Code session in balance. A band above the prompt shows the branch and who started it, open PRs or MRs per person, the branch's CI (passed, failed with the job, running), RAM and disk with warnings, the context level (fresh, mid, full), the running agent and its task, the current task step and the session cost; /mizan opens the details. At full context it saves a hafiz handoff and puts /clear in the prompt for you to send. English and Arabic.

by Loai Elattar

Nexika prof: turns Claude into a personal programming professor: teaches concepts step by step, explains codebases to juniors joining a project, quizzes understanding, and remembers each learner's progress.

by Loai Elattar

Nexika siyaq: project knowledge loaded only when relevant. Entries generated from your docs (no hand-written index), multilingual matching (Arabic included) with relevance ranking and a token budget, file-based triggers, and stats that show what helped and what is missing.

by Loai Elattar

Nexika tabib: finds out why CI failed, with evidence. Reads the failed run (GitHub Actions or GitLab CI), pulls out the failing tests and errors, sorts the failure (code, one Python or OS only, flaky, infrastructure, dependencies), compares with the last green run, reproduces the failing tests in a throwaway git worktree, and reports the cause with its evidence. Diagnoses only: it never edits code, pushes or re-runs CI. English and Arabic.