Index of every page on this site: llms.txt
All skills
hktitan avatar

/khe-auto

@945e23c
by Harshit Khemanihktitan/skills47 stars
2
Run it yourself

The mode that runs the rest. Reads a request, picks a playbook or the one or two skills that own it, says why in a line, then does the work in order. Covers every skill in this collection: the khe skills (UI craft, design tokens, SVG, diagrams, sound, launch films), the skills by Emil Kowalski and Benji Taylor linked beside them, Vercel's `ui-sdk` and agent skills, and the engineering skills of Lauren Tan (poteto) and Matt Pocock. Use when a request names no skill, spans several, or someone says "auto", "which skill", "where do I start", "what order", "list the skills", "make this better", "I need a launch video", "build this screen", "build it with ui-sdk", "we ship Thursday", "the agent keeps doing this". When the request already names a skill, use that skill.

Use this Skill: https://skilld.dev/gh/hktitan/skills/khe-auto

Nothing lands on disk. Nothing to clean up.

Fork this Skill

Edit a local copy. It keeps the author and licence.

README.md

≈1.7k tokens on demand. Your agent reads this file only when SKILL.md points to it.

khe-auto

Picks the right skill for a request and runs it.

npx skills add HKTITAN/skills --skill khe-auto

What it does

  • Reads a request and names the one or two skills that own it, with a one-line reason each.
  • Runs them in order, one at a time, looking at the result between them.
  • Routes to the khe skills, to Emil Kowalski's and Benji Taylor's skills and to Vercel's ui-sdk and agent skills, and to Lauren Tan's (poteto) and Matt Pocock's engineering skills by name, and says whose each one is.
  • Routes by five domains: Design, Code, Brains, Content and Engineering. Engineering covers how agents work in a repo, starting with khe-poteto-pyramid.
  • Sends "show this off" jobs to khe-package: a clip of one thing you built, made for people who have never heard of it.
  • Says so when no skill applies, instead of stretching one over the wrong problem.
  • Runs one loop for whichever skill it routed to: plan, start from gold, build, run that skill's gate, fix the largest fault, walk verify, fresh-eye review, hand over.
  • Places every correction made during the job with khe-poteto-pyramid before handing over: the fix goes to the highest layer that can hold it, and the record goes in the project's PYRAMID.md.
  • Ends every reply with what it could not verify.

How it runs

khe-auto's flow: a request becomes a goal line, then a route to a playbook or to one or two skills by domain; each skill runs its own loop and gate, every correction is placed with khe-poteto-pyramid, the reply passes route-check, and the job is handed over with ROUTE.md

Every request lands in one of five domains. Most of the skills are linked from their authors' repos rather than kept here; the counts are the rows in SKILL.md's domain tables.

The five domains and how many skills each routes to: Design 9 here and 22 linked, Code 12 linked, Brains 7 here, Content 1 here, Engineering 1 here and 34 linked

Used

A request in the words of the playbook table:

/khe-auto I need a launch video for Friday.

The reply for it is the gold in examples/playbook.txt, the bar every playbook reply is held to. It was written by hand as the standard, not logged from one session:

Goal: A launch film for Friday's release that can be posted as it is; I hand back the render and its measured numbers.
Route: playbook `launch-film`
Work:
1. Register: named and written down with the runtime, by `khe-launch-film`. No gate, a decision.
2. Beats: a beat sheet with a time for the first word, the name and the endcard. `khe-launch-film/scripts/film-check.mjs --gate` on the sheet, last line pasted in the handover file.
3. Build: rendered in the project's film engine from the closest gold beat sheet. Same gate, run on the render.
4. Lock picture: no frame will move again. Same gate, run on the locked cut.
5. Sound: one family, a cue sheet with a frame number per sound, by `khe-sound`. `khe-sound/scripts/sound-check.mjs --gate` on the cue sheet.
6. Review: cuts per minute, still share and loudness measured on the file and listed in the handover file. Both gates run on the final file.
- Step 4 was reopened once: a cut landed off its beat. Fixed in step 4, then steps 5 and 6 run again from the top.
- Fresh-eye review saw the render and the gold only, and returned "nothing".
Not verified: loudness on phone speakers, and the vertical cut, which was not rendered.

node scripts/route-check.mjs examples/playbook.txt --gate on that file ends route-check passed, with 0 never-intentional lines and 0 candidates.

When to use it

Use it when you do not know which skill covers a problem, when two seem to overlap, when a job spans several, or when you just want the right thing to happen without choosing. If you already know the skill, call that skill directly: it is one step shorter.

Invoke with

/khe-auto [request | which <job> | list]

Or describe the problem in your own words.

Before and after

Before: "make the logo reveal better" loads four skills at once and returns a list that mixes path cleanup, easing, sound and export settings, none of it applied.

After: a Goal: line, then khe-svg: the reveal lives inside the file, then the svg pass with its changes made and its gate's last line, then Not verified: and the one thing nobody looked at.

Example prompts

  • I need a launch video for Friday. What do I run, and in what order?
  • Which skill do I use for a hero demo of the app on the landing page?
  • The drawer feels off when I let go of it. Where do I start?

Verify

node scripts/route-check.mjs reply.txt --gate   # a routing reply, saved as text
node scripts/route-check.mjs playbooks --gate   # every playbook
node scripts/route-check.mjs --self-test        # the gate against its own fixtures
node scripts/gate-all.mjs                       # every gate in the collection

route-check.mjs names every fault as a line. --gate ends route-check passed only when every never-intentional line is at 0, exits 1 with the counts otherwise, and exits 2 with inconclusive when it could not look.

What it proves: the reply opens with the goal, names two skills at most or one playbook, names only skills in the domain tables of SKILL.md, puts the contract first and picture before sound, gives no design value in the route, gives the install line for a missing skill, and ends with what was not verified. For a playbook: every step names a real skill, an exit and the gate that checks it, and the file ends with its required reply.

What it does not prove: that the route is the right one for the request, that the work followed the skill's rules, or that any gate named in the reply was run. It reads text and never sees the request. Passing is necessary, not sufficient.

gate-all.mjs runs the --self-test of every gate installed beside this skill and prints one line per skill: ok, FAIL, no gate or not installed. A missing sibling is reported and is never counted as a pass.

What is inside

  • SKILL.md: the one idea, the workflow, the playbook table, the domain tables, the overlaps and the output format.
  • playbooks/: nine whole jobs. Every step names its skill, its exit and its gate.
  • references/: loop.md (the shared loop), verify.md (the last pass and every gate line), review.md (the fresh-eye brief), teaching.md (the pairs), and gotchas.md and pov.md, which are yours to edit.
  • examples/: six gold routing replies: one skill, two skills, a playbook, nothing fits, not installed, and which.
  • teaching/: fourteen bad/good pairs, each named for the gate line its bad half raises. They are the gate's tests.
  • scripts/: route-check.mjs (the gate) and gate-all.mjs (every gate's self-test). Node 20 or newer, no dependencies.
  • assets/: the two figures above, flow.svg and domains.svg.
  • LICENSE: MIT.

Source: SKILL.md on GitHub

skilld matched fixed text patterns in SKILL.md and file names. Patterns miss obfuscated code.

skilld run checks every file with the same patterns. It asks for approval before it loads a Skill with a behavior marked Needs approval.

1 warning4d3 checks · Risk SAFE
  • Gen Agent Trust Hub4d

    The khe-auto skill is an orchestrator that manages a collection of specialized design and engineering tools. It analyzes user requests to determine the correct workflow and coordinates the execution of other skills. It features a robust internal verification system for testing installed components and follows strict output and verification protocols.

  • Socket4d

    No alerts

  • Snyk4d

    Risk: MEDIUM · 1 issue

Signed by skilld at 945e23c. This ties the file your Agent reads to that commit on GitHub. It does not review the instructions.

Last checked against GitHub yesterday.

Activeupdated 3 days ago
metadata
{
  "author": "Harshit Khemani",
  "version": "1.1.0"
}

README badge

README badge for hktitan/skills/khe-auto