Skill creator

Skill vindreshsingh/engineering-skills/skills/skill-creator

The broadest agent-skills library: 64 behaviorally-tested engineering skills + a 38-role SDLC & marketing org + an orchestrated build loop. Works with Claude Code, Cursor, Gemini, Copilot, Codex.

Install

npx -y skills add vindreshsingh/engineering-skills --skill skill-creator

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Authors a new engineering skill (or fixes an existing one) so it fires at the right time and gets followed. Use when adding a skill to this repo, splitting an overloaded one, or improving a skill that isn't being triggered.

SKILL.md

3.8 KB, as published. Nobody here has run it

Skill Creator

A skill is only valuable if an agent loads it at the right moment and then actually follows it. Most weak skills fail on one of those two: a vague description that never triggers, or "advice" with no executable process. This skill is the process for writing skills that don't fail that way — it dogfoods this repo's own anatomy.

When to Use

Adding a new skill to this repository
A skill exists but rarely fires, or the agent ignores its steps
One skill is trying to cover several jobs and should be split
Reviewing a contributed skill before merge

Process

Confirm it's process-worthy. A skill encodes a repeatable workflow where people predictably cut corners. If it's a one-off fact or generic advice, it's a doc or a reference — not a skill.
Check it doesn't already exist. Search existing skills; extend or link rather than duplicate. Overlap dilutes routing.
Name it short, kebab-case, action-oriented; the directory name must equal the frontmatter name.
Write the description last and carefully — it's the trigger. Start with what the skill does (third person), then concrete conditions ("Use when…"). Name the situations and keywords an agent would see. This single line decides whether the skill ever loads.
Write an executable Process. Numbered steps, a decision or action in each, ordered by what matters most. If it can't be followed step-by-step, rewrite it until it can. Add short illustrative examples only where they clarify.
Write the anti-shortcut sections. Common Rationalizations (the excuse → the rebuttal) and Red Flags (observable signs it's being violated). These are what make the skill stick under pressure — don't skip them.
Make Verification a checkbox list of observable, testable exit conditions.
Link, don't duplicate related skills with [[name]]. Keep it tight — one screen of signal.
Validate and catalog. Run scripts/validate.sh and scripts/generate-catalog.sh; fix anything that fails ([[pipeline-ops]] enforces this in CI).
Test the trigger and the value. Does an agent load it for the intended task? Does following it produce a better result than not? Iterate the description and steps from real runs.

Common Rationalizations

"The description is fine, the content is what matters." — A skill that never triggers is never read; the description is the product.
"It's mostly advice but useful." — Advice doesn't change behavior under pressure; give a process or don't ship it.
"Rationalizations and red flags are filler." — They're exactly what stops the steps from being skipped.
"I'll test it later." — An untriggered or unfollowed skill is dead weight; verify it works.

Red Flags

A description with no concrete trigger conditions
A "Process" that's a list of principles, not actions
Missing Common Rationalizations, Red Flags, or a testable Verification
A skill overlapping an existing one instead of linking it
name not matching the directory; validation not run
Never tested against a real task

Verification

The topic is a repeatable process people cut corners on (not advice/one-off fact)
No existing skill already covers it; related skills are linked
description leads with what it does, then specific triggers
Process is numbered, executable, and prioritized
Common Rationalizations, Red Flags, and a checkbox Verification are present
scripts/validate.sh passes and the catalog is regenerated
Triggering and usefulness checked on a real task

Gives 0 of the 12 instructions most skill authoring skills give

Counted across 521 of the 523 authors here whose files we hold, read 2026-08-06

keep skill files under 500 linesin 182 of 521, across 89 files
use imperative form in instructionsin 101 of 521, across 30 files
draft assertions while test runs are in progressin 88 of 521, across 22 files
save test cases to evals jsonin 87 of 521, across 21 files
create two to three realistic test promptsin 85 of 521, across 20 files
write skill descriptions to be pushyin 84 of 521, across 19 files
ask questions about edge cases and input formatsin 81 of 521, across 16 files
save timing data immediately when runs completein 74 of 521, across 9 files
include all trigger conditions in the skill descriptionin 73 of 521, across 7 files
capture intent before writing a skillin 70 of 521, across 4 files
launch all test runs in a single turnin 68 of 521, across 2 files
write the description in third personin 56 of 521, across 19 files

Said here and by no other author read

Name it short, kebab-case, action-oriented
Write a numbered, executable process
Write common rationalizations and red flags sections
Make verification a checkbox list of testable exit conditions
Link related skills instead of duplicating them
Test the trigger and value on a real task

Grouped from the skills themselves: near-identical wordings counted once, and counted by distinct author, so one author publishing three of these counts once.