Tests, commits, PR descriptions, docstrings, .env.example, Dockerfile audit. Drop the folder in .claude/skills/, type your task, consistent output every time.
7-day no-questions refund · Team $49 · Developer $79 (client-deliverable)
"Write tests for cart_store.py"
Monday → full pytest file, parametrize, fixtures
Tuesday → 3 stubs, "would you like me to add more?"
Wednesday → integration tests you didn't ask for
"Write tests for cart_store.py"
Monday → complete file, fixtures, parametrize, pytest.raises, ≥70% branch coverage
Tuesday → same
Wednesday → same
A skill is a SKILL.md file with frontmatter + procedure + output constraints. Claude Code reads the frontmatter description, matches your task, injects the procedure. Output locks in.
Triggers on: "write tests for X", "add pytest coverage"
Reads the source, maps public callables, writes complete test file. Happy path + edge + error, fixtures, parametrize, pytest.raises. ≥70% branch coverage. No TODOs.
Triggers on: "commit this", "write commit message"
Reads staged diff, classifies type (feat/fix/refactor/…), picks scope, writes imperative subject ≤72 chars + body that explains WHY. Closes #N footers.
Triggers on: "write PR description", gh pr create
Reads git log <base>..HEAD, produces Summary + What changed + Test plan (checkboxes) + Breaking + Follow-ups. Grouped by subsystem.
Triggers on: "add docstrings", "document this file"
Inserts Google-style docstrings on every public callable. Args/Returns/Raises from signature + body. Only sections that have something to say.
Triggers on: "generate .env.example", "what env vars does this need"
Greps os.getenv, process.env, os.environ[] across the codebase. Groups by concern. One-line comment per var pointing at source.
Triggers on: "optimize Dockerfile", "my image is too big"
Audits size (base, multi-stage, layer order), build cache, security (non-root, no secrets, HEALTHCHECK), runtime (exec CMD). Findings with 🔴/🟡 + line numbers + revised Dockerfile.
mkdir -p .claude/skills && cp -r skills/* .claude/skills/
# macOS / Linux
cp -r skills/* ~/.claude/skills/
# Windows (PowerShell)
xcopy /E /I skills "$env:USERPROFILE\.claude\skills\"
# In Claude Code:
/skills
# Expected: all 6 names listed
Restart Claude Code after install. File watcher picks up skills on new sessions, not mid-session.
Bundle with the Telegram Order Bot, Claude Dev Prompts, and Python Automation Pack — code BUNDLE25 at checkout for $7 off each.
Current Nov 2026+. Skills format SKILL.md + frontmatter is stable in Claude Code 2.0+. On earlier builds without skill support, paste a SKILL.md body into chat as a prompt — same output, no auto-match.
Auto-trigger is a Claude Code feature. On other clients the SKILL.md bodies work as high-quality standalone prompts — copy the Procedure + Output constraints sections, paste with your task.
v1.0 ships 6 skills, four of which are language-agnostic (commit, PR, env, Docker) and two Python-specific (tests, docstrings). v1.1 (Q1 2027) adds Rust test generator, Go error-wrapping, TypeScript React scaffolder — free for existing buyers.
Yes. Fork for your team's conventions, rewrite procedures, tighten constraints. The license restricts only reselling the raw pack as a product.
Personal = you + your projects + your employer (one seat). Team = share with up to 10 teammates at one company. Developer = include in client deliverables (set up a client's Claude Code and ship the skills as part of handoff). Pays back on the first $80 client setup.
7-day no-questions through Gumroad.
You already use Claude Code. These are the skills that make the output stop drifting.
Get the pack — $197-day refund · Instant download · Free updates through Nov 2027