port-parity-testing
Skill2 starsChanged 18 days agoMITView on GitHub
Use when porting code with parity tests against a reference implementation - what parity pins, which oracles are legal, fixture discipline, and differential testing in containers.
In this repository40 other files it gives its agents.
- android-debloatSkill
- arch-linux-cachyosSkill
- beansSkill
- chezmoi-workflowsSkill
- coding-rulesSkill
- communicationSkill
- context7Skill
- docker-compose-setupSkill
- domain-modelingSkill
- audioSkill
- deploymentSkill
- ecsSkill
- game-loopSkill
- inputSkill
- physicsSkill
- renderingSkill
- resourcesSkill
- ebitenSkill
- ebook-convertSkill
- email-inbox-triageSkill
- gitSkill
- go-browser-testsSkill
- go-devSkill
- go-htmx-frontendSkill
- go-setupSkill
- go-testsSkill
- grillingSkill
- interactive-browser-testingSkill
- keep-on-goingSkill
- mcpcSkill
- obsidian-vault-triageSkill
- pixellabSkill
- playwright-mcpSkill
- public-repo-maturitySkill
- researchSkill
- retroSkill
- shell-startup-filesSkill
- shell-trapsSkill
- skill-craftSkill
- web-admin-uiSkill
The file
We split this skill into the rules it gives an agent
2 rules. About 0 are common, 2 are distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈2
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 26
Read the upstream ENCODER's emission table to define the opcode set: the decoder alone omits the cases the encoder never produces, and the …
- Distinctive: line 35
Commit fixture bytes and their manifest: env-gated upstream reads make CI vacuously green (no fixtures means no comparisons, which passes).
≈ “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
Port parity testing
Parity pins BEHAVIOR, never SHAPE: fixture shapes, naming idioms, and signature-mirror functions that exist only to look like the upstream are violations. The port answers the same inputs with the same outputs; it owes nothing to upstream's file layout.
Oracles
- An oracle must not re-derive the implementation's arithmetic: copying the port's formula into the test makes the test pass when the port is wrong twice the same way. Tables carry literals; a differential oracle exists only in the randomized layer, where port and reference run side by side on generated inputs.
- Read the upstream ENCODER's emission table to define the opcode set: the decoder alone omits the cases the encoder never produces, and the parity corpus inherits that blind spot.
- Derive checker semantics from the upstream reference FILE, never from the port's own docs: docs describe intent, the reference defines behavior.
Fixtures
- Commit fixture bytes and their manifest: env-gated upstream reads make CI vacuously green (no fixtures means no comparisons, which passes).
- Count lines with
awk 'END{print NR}', notwc -l: a file without a trailing newline undercounts, and a count assertion built on it passes on truncated fixtures. - Golden digests prove nothing about opcode boundaries the corpus never reaches: hand-craft frames exercising both sides of each boundary.
- Output-orientation parity hides from digests: same digest, different internal state. Pin it with asymmetric fixtures (inputs that differ, outputs that must match).
Differential runs
- Scripted RNG tapes with independent line-by-line traces: when the runs diverge, the trace names the exact line, and a shared transcript cannot.
- In containers: a golden sequence, a seeded lockstep reference, and edge rows run in one compose file so both images see the same bytes.
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.

