agentic-forges/grade-openspec
Watch
4
0
Fork
You've already forked grade-openspec
 0
Code Issues Pull requests Activity
No description
13 commits 1 branch 0 tags 81 KiB
  • Python 57.2%
  • Shell 42.8%
Find a file
Exact
Repository files (latest commit first)
Filename Latest commit message Latest commit date
Christoph Görn cfe92779f8
 chore: 🔧 move authoritative home to b4mad Forgejo 
Migrated from codeberg.org/goern/grade-openspec to
git.b4mad.industries/agentic-forges/grade-openspec.

- repoint README, AGENTS.md, plugin.json and .beads sync.remote
- drop .forgejo/workflows/ci.yml — no runner on the new host, no CI
- the GitHub copy is a stale snapshot now; no push mirror exists
2026-07-30 21:24:44 +02:00
.beads chore: 🔧 move authoritative home to b4mad Forgejo 2026-07-30 21:24:44 +02:00
.claude docs: 📝 add Lage.Bonn image generation prompt template 2026-06-13 13:52:02 +02:00
.claude-plugin chore: 🔧 move authoritative home to b4mad Forgejo 2026-07-30 21:24:44 +02:00
scripts refactor: 🏗️ move skill from .claude/skills to top-level skills/ 2026-06-13 13:29:46 +02:00
skills/grade-spec refactor: 🏗️ move skill from .claude/skills to top-level skills/ 2026-06-13 13:29:46 +02:00
.gitignore bd init: initialize beads issue tracking 2026-06-13 13:48:07 +02:00
AGENTS.md chore: 🔧 move authoritative home to b4mad Forgejo 2026-07-30 21:24:44 +02:00
CLAUDE.md bd init: initialize beads issue tracking 2026-06-13 13:48:07 +02:00
LICENSE chore: initial commit 2026-06-13 12:34:28 +02:00
README.md chore: 🔧 move authoritative home to b4mad Forgejo 2026-07-30 21:24:44 +02:00

README.md

grade-openspec

skills.sh

A Claude Code skill that grades a completed OpenSpec change using the Spec Quality Index (SQI) — a 0–100 score measuring how cheaply the spec implemented, not whether the feature or code is good.

What it measures

SQI answers two questions independently:

Sub-index Question
Implementability (I) Could a competent implementer build this without help?
Fidelity (F) Did the spec match reality?

Every place implementation deviated from, resolved, or corrected the spec is a divergence, classified S0 (clean) through S7 (blocker). Penalty weights penalize escalations and blockers most heavily — needing a human is the headline failure.

Grade bands: A ≥90 · B ≥80 · C ≥70 · D ≥60 · F <60

Full methodology: skills/grade-spec/references/spec-quality-index.md

Install

npx skills add goern/grade-openspec

The short form resolves via the GitHub copy (where skills.sh indexes installs), which is no longer synced. The authoritative source lives on b4mad Forgejo — see Repository.

Usage

After applying or archiving an OpenSpec change:

/grade-spec <change-name>

The skill reads the spec, implementation diff, and run journal; classifies every divergence; and emits the SQI JSON with both sub-scores, a confidence level, and concrete write-back recommendations for the next spec.

Output

Two durable artifacts per run:

  • <change-dir>/sqi.md — human-readable report with divergence table and write-back recommendations
  • ./sqi.jsonl — machine-comparable ledger (one JSON line per grading run)

Repository

Open issues and PRs on b4mad Forgejo.

License

GPL-3.0-or-later