LEVEL 02 — OPEN-SOURCE AI SKILLS
11 Mile Labs Skills
A multi-harness library of small, reusable AI agent skills for Claude Code, Codex, and Gemini CLI — the craft of good engineering, QA, product, and work management, packaged so an agent can pick the right move.
● PUBLIC SOURCE LIVEOpen sourceAI skillsMulti-harnessMIT
▸ BRIEFING
Coding agents are capable but unguided — they will happily do the wrong thing well. The difference between a useful run and a wasted one is usually craft: scoping the work, mapping risk, pressure-testing a plan before acting.
11 Mile Labs Skills encodes that craft as small, portable units an agent can reach for. It is not an installer, not a plugin, and not a vertical business pack — just durable, reusable skills that behave the same in every harness.
▸ WHAT IS INSIDE
ENGINEERING · 6 SKILLS
dependency-upgrade-risk-check
Classify upgrade risk and define what to check before updating a dependency.
deprecation-plan-brief
Plan safe removals with compatibility, notice, and rollback.
implementation-drift-check
Check changed files against stated intent to catch scope drift and stray edits.
regression-risk-map
Map likely regressions and the targeted checks for a specific change.
release-readiness-brief
Decide whether a change is ready to release, needs more work, or should be held.
scope-check
Audit scope creep, hidden complexity, and v1/v2 boundaries.
QA · 5 SKILLS
bug-repro-brief
Turn messy bug reports into reproducible briefs before debugging or ticketing.
bugfix-retrospective-note
Capture a bugfix's cause, fix, prevention, and follow-up without ceremony.
test-failure-signal-brief
Classify a failing test and its output before debugging or rewriting assertions.
test-strategy-slicer
Choose focused test layers for a change instead of coverage theater.
verification-contract
Define what must be checked before calling a feature or ticket done.
PRODUCT · 5 SKILLS
acceptance-criteria-tightener
Rewrite vague acceptance criteria into observable pass/fail statements.
decision-record-snapshot
Capture lightweight decision records with rationale, scope, and revisit triggers.
devil-advocate
Argue against a plan in good faith before committing.
explain-like-5
Force a simple explanation of a concept, architecture, or decision.
pre-mortem
Imagine a project failed and work backward to surface the risks.
WORK MANAGEMENT · 4 SKILLS
handoff-readiness-brief
Make active work transferable before pausing, resuming, or handing off.
review-comment-action-plan
Turn review comments into an ordered action plan with clarifications and deferrals.
ticket-readiness-audit
Check whether a ticket is ready to claim, clarify, split, block, or send back.
work-slice-splitter
Split broad work into independently claimable slices before execution.
▸ WHAT I OWNED
- Authored the skill taxonomy — 20 skills across four disciplines: engineering, QA, product, and work management.
- Wrote each as a standard SKILL.md so it runs identically in Claude Code, Codex, and Gemini CLI.
- Set up the one-command install path: npx skills add 11-Mile-Labs/skills.
- Released it MIT so teams can fork, extend, and inspect every skill.
▸ HOW IT WORKS
WRITE
SKILL.md
standard skill format
skills CLI
npx skills add
ANY
HARNESS
Claude Code · Codex · Gemini
// One format, every harness — skills are content, not lock-in.
▸ DECISIONS THAT MATTERED
Skills, not a plugin
Why: portable content survives harness churn; a plugin would tie the work to one tool.
Standard SKILL.md
Why: a shared format means a skill written once runs everywhere, with no per-harness rewrite.
Discipline taxonomy over a prompt pile
Why: organizing by engineering / QA / product / work makes the right skill findable under pressure.
▸ WHAT SHIPPED
- Twenty reusable skills spanning four disciplines.
- One-command install and multi-harness compatibility out of the box.
- A public MIT repository teams can adopt and extend.
NEXT LEVEL
03 →11 Mile Labs SEO/AEO Skills