abdurshd

@abdurshd source repo

14 skills
0 followers
10 hours ago last updated

▌
Ship · abdurshd
Commits, pushes, and ships the current work with a consistent, safe ritual that surveys the diff, cleans gitignore, uses short commits with no agent attribution, and runs the requested delivery mode. Use when the user says "commit", "push", "ship", "merge to main", "make a PR", requests the ship skill, or asks to address PR review comments. Supports commit-and-push, migrations, mainline merges, pull requests, and review-comment modes.

0
installs
▌
Unslopify · abdurshd bundle
Audits and improves websites or web interfaces that feel AI-generated, generic, template-like, overdecorated, visually homogeneous, or insufficiently crafted. Use for requests to remove AI slop, de-AI a design, make a landing page feel authored, replace generic SaaS aesthetics, remove default icon tiles, standalone card icons, or AI metaphors, reduce card soup or decorative effects, strengthen product-specific art direction, improve UI hierarchy and copy, or perform an anti-slop review before shipping. Supports landing pages, marketing sites, portfolios, docs, dashboards, apps, and individual components; can produce a read-only critique or implement and verify fixes. Uses the open Agent Skills format without requiring a specific agent or harness.

0
installs
▌
Claude Use · abdurshd bundle
Launch Claude Code from Codex as a detached implementation worker after a plan is already agreed. Use when Codex should stay the manager, convert an approved plan into a detailed execution brief, and hand large or parallelizable repo work to Claude Code in headless `claude -p` mode with custom agents, `--model opus`, and bypassed permission prompts. Best for medium/large implementation tasks, UI-heavy work where Claude should drive visual decisions, multi-file refactors, or verification-heavy changes in a local repo. Do not use for trivial edits, pure brainstorming, or tasks that still need open questions resolved first.

0
installs
▌
Codex Grok · abdurshd bundle
Runs the hardened Codex/Grok two-model implementation preset in which OpenAI GPT-5.6 Sol at xhigh plans, orchestrates, and verifies while Grok 4.5 at high performs implementation through bounded Grok Build CLI workers. Use when the user requests codex-grok, asks Codex to supervise Grok, or needs this exact provider-specific safety wrapper. For arbitrary orchestrator or worker tools, use the portable fable-opus protocol instead.

0
installs
▌
Fable Opus · abdurshd
Orchestrates large implementation plans by keeping the active agent focused on planning, delegation, and verification while isolated worker agents perform all code changes. Defaults to Fable 5 as orchestrator and Claude Opus 5 through the latest Opus model selector as implementer, but honors any host, worker tool, or model explicitly selected by the user. Use for delegated implementation, parallel workstreams with disjoint ownership, orchestrator-worker pipelines, or requests to implement through Opus, Claude, Codex, Cursor, or another coding agent.

0
installs
▌
I18N Sweep · abdurshd
Sweeps a codebase for internationalization gaps, including hardcoded user-facing strings, missing translation keys, and locales that drifted out of sync, then fills the gaps with proper translations and verifies the result. Use when UI is untranslated, screenshots show mixed languages, the user asks to add a language, or an i18n audit is requested for a page, directory, or whole project.

0
installs
▌
Codex Review · abdurshd
Iteratively reviews and improves an implementation plan with an independent coding agent. Defaults to OpenAI Codex CLI using GPT-5.6 Sol with high reasoning and a read-only sandbox, but honors any reviewer tool, model, or harness explicitly selected by the user. Use for second-opinion plan reviews, adversarial risk checks, iterative approval gates, or requests to have Codex, Claude, Cursor, or another agent critique a plan before implementation.

0
installs
▌
Fable Review · abdurshd bundle
Iteratively reviews and improves an implementation plan with an independent coding agent. Defaults to Claude Code CLI using its latest Fable model alias at high effort and read-only plan mode, but honors any reviewer tool, model, or harness explicitly selected by the user. Use for second-opinion plan reviews, adversarial risk checks, iterative approval gates, or requests to have Fable, Claude, Codex, Cursor, or another agent critique a plan before implementation.

0
installs
▌
Fix Findings · abdurshd
Processes a pasted findings list from an external reviewer, PR review bot, security audit, or coding agent by verifying every claim against current code, fixing confirmed defects, rebutting stale or intentional findings, and reporting per-finding outcomes. Use when the user pastes P1/P2-style findings, says an agent found issues, asks to fix a review, or requests the fix-findings skill.

0
installs
▌
Genvid Promo · abdurshd bundle
Create polished Remotion product-promotion videos from a repository, live product, design files, screenshots, or a product brief. Use when asked for a launch video, hero video, feature promo, product teaser, social promo, or narrated product showcase that must use verified product UI, components, tokens, transitions, animation, synchronized TTS voiceover, and a rendered MP4.

0
installs
▌
Genvid Tutor · abdurshd bundle
Create complete narrated teaching videos for any topic using Remotion, including lesson design, accurate research, animated explanations, diagrams, examples, synchronized TTS voiceover, captions, visual QA, and final MP4 rendering. Use for educational explainers, concept lessons, course segments, worked examples, training videos, or animated tutorials that are not necessarily tied to a software product.

0
installs
▌
Claude Review · abdurshd bundle
Use Claude Opus 5 through the authenticated Claude Code CLI as a detached critical reviewer for either (1) implementation-plan review before coding or (2) uncommitted-code review before commit. Trigger this skill when Codex already has a concrete task and either needs an independent Claude review to stress-test the plan against the codebase, security, and correctness concerns, or needs Claude to review local workspace changes for correctness, regressions, security, and alignment with the target task. Best for medium/large tasks, security-sensitive changes, refactors, and any work where Codex should iterate with Claude until the plan or implementation is defensible.

0
installs
▌
Genvid Onboard · abdurshd bundle
Create detailed step-by-step onboarding and product-demo videos for a specific website, app, repository, project, or workflow using verified real UI and behavior. Use when asked to explain how a product works, demonstrate setup or a user journey, record an onboarding walkthrough, or teach exact product usage with Remotion animations, cursor actions, synchronized TTS narration, captions, and a rendered MP4.

0
installs
▌
Fable Opus Codex · abdurshd
Runs a delegated implementation pipeline followed by an independent iterative code-review gate. Defaults to Fable 5 orchestration, Claude Opus 5 through the latest Opus model selector for implementation, and OpenAI Codex CLI review, but honors any host, implementer, reviewer tool, or model explicitly selected by the user. Use for large multi-agent implementations that require disjoint ownership, orchestrator verification, and a second-agent approval verdict before shipping.

0
installs

14 published skills