45 production Claude Code, Codex and Gemini skills, free to download
Every file on this page is a working SKILL.md exported from the agent tree that runs Bles Software. Download one, drop it into ~/.claude/skills, ~/.codex/skills, or a Gemini project instruction file, and the agent picks it up on the next run.
Agent Engineering skills for Claude Code and Codex
Skill Creator
Write a new Claude Code skill the way Anthropic's own guidance says to write one: sharp trigger, tight scope, real examples.
Use when a repeated workflow deserves to become a skill instead of a prompt you retype.
Agent Creator
Scaffold a Claude Code subagent with its own system prompt, tool allowlist, and handoff contract.
Use when a job needs its own context window and its own tool boundary.
Hook Creator
Build Claude Code hooks that fire on tool events, so a rule is enforced by the harness instead of remembered by the model.
Use when you catch yourself saying 'from now on, always do X before Y'.
Plugin Creator
Scaffold a Codex plugin directory with a valid manifest and a personal marketplace entry.
Use when packaging skills, prompts, or agents for distribution to a team.
Claude Sessions
Save, list, and reconnect to previous Claude Code sessions instead of losing a long-running thread.
Use when work spans days and the context is worth more than the transcript.
CLI Orchestrator Fallback
Call Claude Code and Codex from Node with primary/fallback routing, schema-validated JSON, and graceful degradation.
Use when an AI CLI sits inside a production service and cannot be allowed to hang the request.
Code Quality skills for Claude Code and Codex
Ponytail
Force the laziest solution that actually works: does it need to exist, is it already here, does the stdlib do it, can it be one line.
Use on any task where the real risk is over-engineering, not under-building.
Long Horizon Dev
A four-file planning system (Prompt, Plan, Implement, Documentation) that gives a project durable memory across sessions.
Use for any build large enough that a fresh context would lose the thread.
Iterative Dev Loop
Stop shooting once. Change, run, verify the real result, compare to the previous round, keep only clear wins, stop at a verified plateau.
Use whenever the output has a checkable score and the first attempt is rarely the best one.
Build Plan Doc and Tasks
Turn a rough idea into a written plan plus an ordered, dependency-aware task file a runner can execute.
Use before autonomous execution, so the agent inherits a plan instead of inventing one.
Plan to Code Async
Execute a multi-part task file autonomously, part by part, with self-healing, testing, and progress reporting.
Use when a project is too long for one sitting and each part can be verified on its own.
Eval Framework
LLM-as-judge evaluation built for determinism: binary rubrics, golden sets, and alignment checks on the judge itself.
Use when 'is this good?' has to become a number you can trust twice.
Playwright Tests Without Pipeline Fire
Integration-test an endpoint whose happy path costs minutes and dollars, without ever firing the expensive pipeline.
Use when the only untested code is untested because testing it is expensive.
Debugging skills for Claude Code and Codex
Auto Error Fixer
Watch a log, detect fixable errors, deduplicate by signature, and dispatch an autonomous fix run for each new one.
Use when a long-running service produces the same class of error faster than a human can triage it.
Idempotency Debugging
Find and fix duplicate side effects caused by timeout-plus-retry against a non-idempotent API.
Use when something happened twice and nobody can find the second call.
Cluster DB Safety
Atomic read-modify-write for processes running in cluster mode, where two workers race the same JSON or row.
Use when state is silently lost under concurrency and the code looks correct in isolation.
Multilingual Punctuation Debugging
Trace unwanted punctuation in non-Latin scripts across every layer that touches the string.
Use when Hebrew, Arabic, or CJK text arrives with characters nobody wrote.
Claude Transcript Bloat Fix
Diagnose the Claude CLI 'Killed' / SIGKILL crash caused by bloated session transcripts, and fix it without losing history.
Use when claude dies on startup or on resume and the machine has plenty of free memory.
Single Endpoint to Orchestrator Reroute
Reroute a feature endpoint that returns a partial artifact back through the platform's real orchestration path.
Use when a shortcut endpoint bypassed history, drafts, and approval and now nothing lines up.
Multi-Stage AI Business Constraints
Enforce hard business constraints across an LLM planner, a media generator, and a branded composer without trusting any stage to remember them.
Use when a rule must hold every time and a model is somewhere in the pipeline.
Design skills for Claude Code and Codex
Hallmark
Anti-AI-slop design: build, audit, or redesign a page so it does not read as a template with the defaults left on.
Use on any greenfield page, redesign, or design extracted from a reference URL.
Apple Design
Apple's interface and motion principles translated to the web: springs, interruptible transitions, materials, optical typography.
Use for gesture-driven UI, sheets, drag and swipe, and motion that has to feel physical.
Animation Vocabulary
Reverse-lookup glossary that turns 'the bouncy thing when a popover opens' into the exact term a designer or model will recognise.
Use when you can describe the motion but cannot name it, and the name is what unlocks the prompt.
Reviewing Visual Output
The quality gate for visual work: open the rendered artifact, check crop, hierarchy, spacing, safe zones, RTL behaviour, then revise and look again.
Use after creating or editing any image, PDF, deck, ad, screenshot, or UI.
Hero Loop Animation
Premium seamless-loop hero videos with transparent backgrounds that survive every browser.
Use for landing-page hero motion that must not stutter or show a seam.
Video & Image skills for Claude Code and Codex
Smooth Web Video
Capture video from anything that renders in a browser and prove it is smooth before it ships, by frame index rather than screen recording.
Use for every published video, because a stuttering file probes as perfectly healthy.
Canvas Chromakey Video
Real-time canvas chromakey for transparent video, including iOS Safari where WebM alpha simply fails.
Use when a transparent video has to work on phones, not just on your laptop.
Smart Background Remove
Background removal for image and video using chromakey or LAB Delta E, chosen by whether glow and soft edges have to survive.
Use when a cutout has to keep light spill, hair, or particle effects intact.
Image Gen
Generate and edit raster images with reference control over likeness, palette, and composition.
Use for social creatives, mockups, textures, and cutouts that need to look intentional.
Media Gen
Generate and upscale images and video through a curated registry of best-in-class models instead of one default.
Use when model choice, not prompt wording, is what decides the output quality.
Visual QC with Gemini Vision
Automated visual QC for generated graphics: capture the frame, have a vision model hunt overlaps, cutoffs, font failures, and missing logos, then retry.
Use when creatives render in a batch and no human will look at every one.
Research skills for Claude Code and Codex
Deep Research
Multi-source research with citation tracking, persisted evidence, and a structured report you can audit later.
Use when the answer has to survive being checked, not just sound complete.
AI SEO
Optimise content to be cited by AI answer engines: AEO, GEO, llms.txt, knowledge bundles, agent-readable structure.
Use when the traffic you want now arrives through an AI answer rather than a blue link.
Autoresearch Machine
Turn one asset and one honest number into an overnight optimisation loop: one change per round, score it, keep winners, revert losers.
Use when feedback is fast and objective, and the asset is genuinely yours to edit.
Defuddle
Extract clean markdown from a web page, dropping navigation and clutter, so a read costs a fraction of the tokens.
Use instead of a raw fetch whenever an agent has to read real pages at volume.
Content skills for Claude Code and Codex
Humanizer
Strip the tells of AI writing: inflated symbolism, hedged conclusions, rule-of-three padding, the em-dash habit.
Use on any text a real reader will judge, based on the documented signs of AI writing.
Conversational Mastery
Communication frameworks for outbound messages: specificity, honest stake, mirrored language, one low-burden ask.
Use for client, partner, and team messages where a generic template would cost you the reply.
Markdown to PDF
Markdown to a genuinely well-set PDF, including Hebrew and RTL that keeps code, paths, and tables the right way round.
Use for proposals, reports, and CVs, especially when the document is bilingual.
JSON Canvas
Create and edit JSON Canvas files with nodes, edges, and groups, so an agent can draw a real diagram.
Use for mind maps, flowcharts, and architecture canvases an agent has to author or update.
Obsidian Markdown
Obsidian-flavoured markdown done properly: wikilinks, embeds, callouts, properties, tags.
Use when an agent writes into a vault and the syntax has to be native, not close enough.
Growth skills for Claude Code and Codex
Paid Landing Optimizer
A daily, evidence-first round on a page taking paid traffic: read the analytics, watch real sessions, run one honest test, ship the change.
Use when money is going into a page and nobody has watched a real visitor use it.
Ad Copy Generator
Ad copy written against a specific creative, so the words reinforce the image instead of competing with it.
Use when the creative is chosen and the copy still has to earn the click.
Ad Creative Monitor
Detect creative fatigue before it kills return: CTR decay, frequency creep, CPC inflation, tracked per ad.
Use when spend is steady and performance is quietly sliding.
Budget Optimizer
Analyse spend efficiency across campaigns and ad sets, then move budget from losers to winners with the reasoning attached.
Use on any account where budget was set once and never revisited.
Dedup Before Email Send
A hard guard at the final second before any email send, blocking a duplicate that another path or lane already sent.
Use on every automated send, because the second copy is the one that costs you the relationship.