v1.1.0

What's new in v1.1.0

Added

• Experience-level diagnostic at Stage 0a so the tutorial right-sizes itself - skips Fundamentals for users who already know them, jumps straight to refinement for users who've already shipped one.
• Demo Mode (Stage 0b): runs the full tutorial flow against a canned release-notes-writer example so users can watch a skill being built end-to-end before tackling their own.
• End-of-tutorial deliverable (Stage 10): the agent now builds a real installable .zip and prints absolute install paths for both Claude.ai upload and cp -r ~/.claude/skills/.
• New reference chapter references/scripts.md: when to bundle code, invocation patterns from SKILL.md, exit-code discipline, security considerations, and a worked Python example.
• New reference chapter references/assets.md: the references/ vs assets/ boundary, template discipline with named placeholders, what does not belong in assets/, and a canonical multi-file skill layout.
• New reference chapter references/evals.md: a 5-prompt starter eval set (3 should-trigger, 2 should-not) for regression protection on top of the three lightweight tests, with manual / semi-automated / programmatic runner options.
• Optional Stage 7b for co-authoring an eval set with the user before they ship.
• GitHub Actions release workflow (.github/workflows/release.yml) that validates the skill structure, builds the installable zip, generates release notes from this changelog, and publishes a GitHub Release on every v* tag.
• This CHANGELOG.md, in Keep a Changelog format.

Changed

• SKILL.md operating flow expanded from 9 stages to 10 (Stage 10 ships the deliverable).
• README What's in the box annotates each chapter with a one-line scope, and a new "What's new" section calls out v1.1 additions.
• README License section clarifies that the MIT license covers this packaging only - the underlying conceptual content remains Anthropic's.
• Pre-flight checklist (references/checklist.md) gained items for scripts/, assets/, evals/, and a zip-listing sanity check.
• Skill metadata.version bumped to 1.1.0.

Documented

• Each new reference chapter opens with a scope note distinguishing source-derived material from community-added practitioner convention.
• The Python example in scripts.md is now line-commented to teach the design choices it embodies (exit-code contract, argparse vs sys.argv, stderr/stdout separation, Path() normalization, sys.exit propagation).

Commits

Changes since v1.0.0:

• Add release workflow and CHANGELOG (82cd953)
• Document new chapters for public consumption (d40c696)
• Add diagnostic, demo mode, end-of-tutorial zip, and scripts/assets/evals chapters (713f74e)
• Add 'See it in action' section to README (8332d7d)

Install

Claude.ai (web or desktop)

1. Download build-a-skill-v1.1.0.zip from the Assets section below.
2. Open Claude.ai → Settings → Capabilities → Skills.
3. Click Upload skill and select the downloaded zip.
4. Toggle the skill on.

Claude Code

# Download and extract into the local skills directory.
curl -L -o build-a-skill.zip \
  https://github.com/egalano/build-a-skill/releases/download/v1.1.0/build-a-skill-v1.1.0.zip
unzip build-a-skill.zip -d ~/.claude/skills/

Try it

Start a new conversation and say one of:

• "Walk me through building a skill"
• "Teach me how to make a Claude skill"
• "/build-a-skill"

build-a-skill v1.0.0

Interactive Claude Skill version of Anthropic's "Complete Guide to Building Skills for Claude."

Inspired by Ali Yahya's suggestion (https://x.com/alive_eth/status/2057306868297425062) that guides like this should be skill files you converse with, not PDFs you scroll.

Install

• Claude.ai: download build-a-skill-v1.0.0.zip, Settings → Capabilities → Skills → Upload skill.
• Claude Code: unzip into ~/.claude/skills/ then ask "walk me through building a skill".