LVLme finds the gaps between you and the job, then turns them into steps you can take this week. Then it measures again, so you can watch yourself level up.
Free and open source agent skill, MIT licence. Claude Code, Codex and Cursor.
/plugin install lvlme --marketplace mark-hallam/lvlmeThen say "run LVLme quick" and paste a job link. Honest odds in a couple of minutes.
Every waypoint on the chart above, in full: what Alex did, when, and how far the evidence is trusted.
Seven actions landed, +33.6 in all: fit 37 on Day 1, 70.6 on Day 76. One more is planned.
One loop, run on any job you point it at: honest about where you are today, specific about what moves you up.
LVLme reads the job and your evidence, then scores five factors with the same rules every time. Evidence or it scores zero.
The gap becomes real-world actions: build proof people can check, meet people inside, earn what's missing, get a referral.
Do the work, then measure again. Every scored day is saved with its date, so your climb is a record anyone can check.
Examples use the fictional demo: Alex Example, Acme AI, Globex Labs and Initech Robotics.
Start small. Point it at one posting and see where you really stand before you commit to a full campaign.
It gives you fixes and a brief, never application answers: you write those yourself. Ready for the whole climb? Say "run LVLme".
LVLme quick: Deployment Program Manager, AI Agents at Globex Labs.
Odds today: Low chance of an interview, very low chance of an offer (fit 41 of 100).
The fixes that would raise them most:
If they all land: fit 55, a moderate chance of an interview, a low chance of an offer.
A blind reviewer checked every score above 2. One disagreement counts as 2 until settled.
A score you can't trust is worse than no score. These rules are in the code, and the code is open.
A claim with nothing behind it adds nothing. A line on a CV alone counts as self-reported.
Every piece of evidence is graded, and a factor can never score higher than its evidence allows.
A second AI scores every point above 2 without seeing the first scores. Until they agree, it counts as 2.
It never sends, posts or applies. Every message is a draft you read, change and send yourself.
Your workspace lives on your machine. Your AI provider sees what the agent reads, as with any agent.
Read every rule and every line of the scoring on GitHub. Change it, run it, check it.
Mark is climbing from aviation and safety towards a dream job, and his own Day 1 measured Very low. That's exactly who LVLme is for: people aiming past their usual path who want the honest number, and a way up from it.
Free and open source. Works in Codex and Cursor too.
/plugin install lvlme --marketplace mark-hallam/lvlmeMIT licence. LVLme used to be called Climb, and old links to github.com/mark-hallam/climb still work.