EffortRankWhat a ledger isGalleryGitHub

EffortRank

Two skills for coding agents, one repository.

resume-ledger builds a work ledger: a work history where every line carries a receipt a reader can check in a minute. From it, the agent reshapes the person's résumé for a posting and writes the cover letter. The build runs on the person's machine, sends nothing, and writes to no account. Recipients read [what a ledger is](LEDGER.md) and can verify one with their own agent.

effortrank is the game on top, published later: push the ledger's receipts, signed with the SSH key GitHub already lists for you, and the gallery computes one number, EffortRank, what it would cost to fake the document. The score exists only on the gallery; nobody sends a score. The formula is public: [gallery/tiers.json](gallery/tiers.json).

Neither scores a person's ability. They describe a document's verifiability. Not a hiring tool, not sold to employers.

Install

npx skills add effortrank/effortrank --skill resume-ledger

Then, in your agent: "apply to this posting" or "build my résumé from evidence". It explains what it will read, proposes defaults, asks once, and writes ledger/: ledger.md, resume.pdf, cover-letter.md. You send them. Someone sent you a ledger? "verify this ledger".

Requirements

bash, curl, git; jq for the one-liners. No GitHub token or gh needed. PDF text via pdftotext or macOS. Nothing is installed by the skills.

Layout

  • skills/resume-ledger/ — SKILL.md (the workflow), ledger.sh (the seal), references/ledger-format.md (the fixed format), references/resume-template.html (black and white, serif, prints to A4), references/observations.md
  • game/effortrank/ — the reveal: SKILL.md and effortrank.sh (push, delete). Outside skills/ until the game launches, so installers do not pick it up yet.
  • gallery/ — the receiving end: server.py (signatures, versions, rank, cards, the explainer at /ledger), verify.py (refetches every receipt, drops what fails, with reasons), tier.py + tiers.json (the number), github_facts.py, ghapi.py. Deploy notes in gallery/DEPLOY.md.
  • LEDGER.md — the explainer recipients read. PRIVACY.md — what leaves the machine, and when.
  • tests/ — test_boundary.sh (honest and padded manifests give the same number; every fake, stolen, malformed or unsigned-login receipt is dropped with a reason), test_gallery.sh (push, padded push, stranger key, unsigned second login, signed delete and republish), trigger-phrase evals.

Status

Ledger format ledger/0.1, formula effortrank/0.1. Gallery live on Cloud Run. See [PLAN.md](PLAN.md) for the reasoning and the open questions.