using-loopkit
starsforks
How to install using-loopkit?
archive228/loopkit/using-loopkit
$npx -y skills add Archive228/loopkit --skill using-loopkitUse when starting any conversation in a loopkit-enabled project - establishes how to find and use loopkit's 49 skills, requiring skill invocation before ANY response including clarifying questions.
SKILL.md
| 1 | # Using Loopkit |
| 2 | |
| 3 | <EXTREMELY-IMPORTANT> |
| 4 | If you think there is even a 1% chance a loopkit skill applies to what you are doing, INVOKE it. |
| 5 | |
| 6 | IF A SKILL APPLIES, YOU DO NOT HAVE A CHOICE. YOU MUST USE IT. |
| 7 | |
| 8 | This overrides "just answer quickly" instincts. Not negotiable. |
| 9 | </EXTREMELY-IMPORTANT> |
| 10 | |
| 11 | ## The Rule |
| 12 | |
| 13 | **Invoke relevant skills BEFORE any response or action** — including clarifying questions, exploring the codebase, or reading files. If it turns out wrong for the situation, drop it. |
| 14 | |
| 15 | Then announce "Using [skill] to [purpose]" and follow the skill exactly. If it has a checklist, create a todo per item. |
| 16 | |
| 17 | ## Where the skills live |
| 18 | |
| 19 | Skills are files at `.claude/skills/<name>/SKILL.md`. Each has YAML frontmatter with `name` and `description` (the description is a trigger phrase, not a summary). Load a skill by reading its SKILL.md when its trigger matches your task. |
| 20 | |
| 21 | ## Skill routing (49 skills, 10 tracks) |
| 22 | |
| 23 | | Task shape | First skill | |
| 24 | |---|---| |
| 25 | | "Fix this bug" / test failing / crash | `systematic-debugging`, then `read-the-trace` | |
| 26 | | "It broke between two commits" | `bisect-regression` | |
| 27 | | "Flaky test" | `flaky-hunter` | |
| 28 | | "Add a feature" / write anything new | `spec-first`, then `write-failing-test-first` | |
| 29 | | "Refactor" / dead code / deep nesting | `kill-dead-code`, `simplify`, `reduce-nesting` | |
| 30 | | About to claim done / commit / open PR | `adversarial-verify` + `verification-before-completion` + `self-eval-bias` | |
| 31 | | Review a diff | `adversarial-verify`, `pr-from-diff` | |
| 32 | | Frontend / UI work | `design-system`, `a11y-pass`, `loading-empty-error-states` | |
| 33 | | Security touch | `owasp-review`, `authz-check`, `input-validation`, `secret-scan`, `dependency-audit` | |
| 34 | | Data / SQL / migrations | `sql-review`, `migration-writer`, `schema-diff` | |
| 35 | | Docs / changelog / README | `changelog-from-diff`, `decision-record`, `readme-audit` | |
| 36 | | Git ops | `clean-commits`, `pr-from-diff`, `rebase-safely`, `revert-surgical` | |
| 37 | | Test suite gaps | `coverage-gaps`, `contract-test` | |
| 38 | | Running out of context | `context-budget`, `tool-restraint` | |
| 39 | | Parallel work | `subagent-fanout` | |
| 40 | | Starting a fresh project / major feature | `planner-spec-expand`, then `feature-list-json`, then `init-script-contract` | |
| 41 | | Bootstrapping into an existing multi-session project | `progress-reading-protocol` | |
| 42 | | Entering an implementation sprint | `sprint-contract` | |
| 43 | | Calibrating a reviewer / evaluator | `evaluator-calibration` | |
| 44 | | New Claude/Sonnet/Opus model landed | `harness-stripping` | |
| 45 | |
| 46 | Full list: `ls .claude/skills/`. |
| 47 | |
| 48 | ## Red Flags — STOP and check for a skill |
| 49 | |
| 50 | | Thought | Reality | |
| 51 | |---|---| |
| 52 | | "This is just a simple question" | Questions are tasks. Check first. | |
| 53 | | "Let me explore the codebase first" | Skills tell you HOW to explore. Check first. | |
| 54 | | "I remember this skill" | Skills evolve. Read the current SKILL.md. | |
| 55 | | "The skill is overkill" | Simple things become complex. Use it. | |
| 56 | | "I'll just do this one thing first" | Check BEFORE doing anything. | |
| 57 | | "Tests pass, we're good" | `verification-before-completion` says: run the exact command, read the output, then claim. | |
| 58 | | "I'll do both features while I'm in here" | `single-feature-discipline` says: one per session. Never two. | |
| 59 | | "The reviewer will let this slide" | `self-eval-bias` says: assume it will confidently praise. Calibrate first. | |
| 60 | |
| 61 | ## Priority when multiple skills apply |
| 62 | |
| 63 | Process skills first (spec-first, systematic-debugging, planner-spec-expand, sprint-contract), then implementation skills (design-system, sql-review, etc.), then finishers (adversarial-verify, verification-before-completion, self-eval-bias, clean-commits). |
| 64 | |
| 65 | - "Let's build X" → `planner-spec-expand` → `feature-list-json` → `sprint-contract` → domain skills → `adversarial-verify`. |
| 66 | - "Fix bug Y" → `systematic-debugging` → `read-the-trace` → fix → `verification-before-completion`. |
| 67 | - "Session open in existing project" → `progress-reading-protocol` → `sprint-contract` → work. |
| 68 | |
| 69 | ## Skill release convention |
| 70 | |
| 71 | Every new skill in loopkit ships as a folder with four required files. No exceptions — a skill without these is a draft, not a release. |
| 72 | |
| 73 | ``` |
| 74 | skills/<skill-name>/ |
| 75 | SKILL.md # the skill itself (frontmatter + procedure) |
| 76 | POST.md # ~200-word X-thread-shaped explainer |
| 77 | evidence/ |
| 78 | before.md # verbatim transcript WITHOUT the skill loaded |
| 79 | after.md # same prompt WITH the skill loaded |
| 80 | ``` |
| 81 | |
| 82 | - `SKILL.md` — routed on its frontmatter `description`. Body under ~150 lines. |
| 83 | - `POST.md` — the announcement thread. Copy from `template/POST.md`, fill in every placeholder before publishing. |
| 84 | - `evidence/before.md` + `evidence/after.md` — one real task, both transcripts. The before/after pair is the receipt that the skill actually changes behaviour. If you cannot pr |
People who liked this also viewed
6 picksType
- find-skillsHelps users discover and install agent skills when they ask questions like "how do I do X", "find a skill for X", "is there a skill that can...", or express interest in extending capabilities. This skSkillsJul 2026
- grill-meA relentless interview to sharpen a plan or design.SkillsJul 2026
- vercel-react-best-practicesVercel's official collection of agent skillsSkillsJul 2026
- improve-codebase-architectureScan a codebase for deepening opportunities, present them as a visual HTML report, then grill through whichever one you pick.SkillsJul 2026
- setup-matt-pocock-skillsConfigure this repo for the engineering skills — set up its issue tracker, triage label vocabulary, and domain doc layout. Run once before first use of the other engineering skills.SkillsJul 2026
- triageMove issues and external PRs through a state machine of triage roles — categorise, verify, grill if needed, and write agent-ready briefs.SkillsJul 2026
