parity-check
Skill0 starsChanged 10 days agoMITView on GitHub
Measures how close this build is to the reference product: a feature parity score from the feature matrix (weighted by must, should, could) with the missing list in build order, plus a screenshot layout diff that ignores colour so a rebrand does not count against you. Two standard-library Python tools. Use when the user says "how close are we", "what's missing", "parity", "is it ready", or "diff the screens".
In this repository42 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- .github/copilot-instructions.mdCopilot instructions
- agent-handoverSkill
- brand-sweepSkill
- design-tokensSkill
- experiment-labSkill
- pitch-from-evidenceSkill
- plan-and-buildSkill
- platform-reachSkill
- product-kickoffSkill
- product-reconSkill
- quality-loopSkill
- repository-auditSkill
- review-miningSkill
- review-practiceSkill
- security-auditSkill
- stack-foundationSkill
- systematic-debuggingSkill
- technical-reconSkill
- ux-qualitySkill
- web-qualitySkill
- agent-handoverSkill
- brand-sweepSkill
- design-tokensSkill
- experiment-labSkill
- parity-checkSkill
- pitch-from-evidenceSkill
- plan-and-buildSkill
- platform-reachSkill
- product-kickoffSkill
- product-reconSkill
- quality-loopSkill
- repository-auditSkill
- review-miningSkill
- review-practiceSkill
- security-auditSkill
- stack-foundationSkill
- systematic-debuggingSkill
- technical-reconSkill
- ux-qualitySkill
- web-qualitySkill
The file
We split this skill into the rules it gives an agent
2 rules. About 0 are common, 1 is distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 41
Make sure docs/product/features.csv is current: every row's status is yes, partial (with a note), no, or skip (with a reason).
- Distinctive: line 95
shippable: all must-haves done, feature score 80+, no open blocker or high-severity issue
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
parity-check
Reads docs/product/features.csv (the one matrix, shared with kickoff, recon,
review-mining and make next), plus the two screenshot sets. Writes
reference/parity.md.
Two tools in this folder, both standard-library Python, no installs (commands run from the project root):
python3 .agents/skills/parity-check/parity.py docs/product/features.csv # feature parity + missing list
python3 .agents/skills/parity-check/imgdiff.py reference/screens/S07.png reference/build-screens/S07.png --out reference/diffs/S07.png
python3 .agents/skills/parity-check/imgdiff.py a.png b.png --json > reference/diffs/S07.json # for parity.py --visual
python3 .agents/skills/parity-check/parity.py docs/product/features.csv --visual reference/diffs/*.json --markdown > reference/parity.mdWhat parity means here
Parity means it does the same job, not that it looks the same. The feature score is the number that matters: can a user do everything they could do in the reference product. The layout score checks structure (is the information in the same places, in the same hierarchy), because that is what makes a switcher feel at home. It deliberately ignores colour, because brand-sweep changes every colour on purpose. Do not chase pixel parity with the reference product. Its exact look is its trade dress, and it changes before launch.
Step 1: feature parity
Make sure docs/product/features.csv is current: every row's status is
yes, partial (with a note), no, or skip (with a reason). status is
does this build have it. An old reference/features.csv with a clone
column still scores (deprecated; rename it to status). Then:
python3 .agents/skills/parity-check/parity.py docs/product/features.csvIt weights must 3, should 2, could 1, counts partial as half, leaves out
skip rows and rows you added that the reference product does not have
(source review-mining, or original no), and
prints: the score, must-haves done of total, each area weakest first, and the
missing list in build order. Must-haves not done means not shippable, and it
says so.
Step 2: layout diff
For each key screen, take two screenshots at the same viewport (1440x900
desktop, 390x844 mobile) in the same state (same data shape, same tab
open, logged in the same way). The reference product's come from public pages
or the user's own account, saved in reference/screens/. This build's go in
reference/build-screens/.
python3 .agents/skills/parity-check/imgdiff.py reference/screens/S07.png reference/build-screens/S07.png --out reference/diffs/S07.pngLayout mode (default) turns both into edge maps, cuts them into a grid, and compares where things are, scoring only cells with something in them. It reports the score (matches 90+, close 75+, partly 50+, different), the regions that differ (in the reference screenshot's pixel coordinates, biggest first), and a height difference if this build's page is much longer or shorter. Retina and non-retina screenshots compare fine: both are scaled to the same width.
--mode pixel is exact comparison. Use it for your own regressions (this
build today against this build last week), not against the reference.
parity.py --visual averages layout scores only and lists pixel-mode or blank
comparisons as ignored.
Step 3: behaviour diff
Walk each flow in both products and compare what the scores cannot see: clicks to finish the core flow, what happens on errors, what is remembered between visits, what emails arrive. Fewer clicks than the reference is a win. Write each difference as: flow, reference does, build does, fix or keep.
Step 4: the report
reference/parity.md: overall score, feature score, layout score per screen,
missing features in build order, behaviour differences, and a verdict:
- not shippable: any must-have missing, or any open blocker issue
- shippable: all must-haves done, feature score 80+, no open blocker or high-severity issue
- better than the reference: shippable, plus fixes from review-mining. This is the goal. A straight copy has no reason to exist.
Give honest numbers. A build at 62% is at 62%.
Output
reference/parity.md, the diff images in reference/diffs/, and the top five
things to build next. Then the project's issue-backed build loop (AGENTS.md)
for the gaps, or review-mining when parity is there.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

