WEFT_OS/.agents/skills/weft-integration-verification/SKILL.md

55 lines
2.9 KiB
Markdown
Raw Normal View History

docs: add engineering contract, contributor toolkit and templates Add AGENTS.md as the single engineering contract: sources of truth, evidence and proof requirements, ownership, working modes, the reviewed change loop, maintenance isolation, decision authority, dependency and fork rules, the 0.0.x to 0.1.0 release progression, and branch and publication rules. Add .agents/ with scope, implement, verify, maintenance and release procedures; reviewer, architecture debt scout, simplicity and refactoring roles; skills for the compositor, Servo rendering, runtime capabilities, appd sessions, packages and data, Linux deployment, integration verification and documentation; and verify.py, which runs the existing check profiles with --locked from any directory, stops on the first failure and supports --dry-run. check_toolkit.py validates local links and heading anchors, metadata, catalog coverage and issue forms. A new contributor-toolkit CI job runs both. The structure follows the Resina Design System toolkit at 1e768064f57a1e403ab271822383e732795905a6, adapted to WEFT. Validation, against the committed lockfiles: - verify.py toolkit: checker passes, 28 runner/checker tests pass - verify.py portable, demos, packages: pass - verify.py linux, runtime, servo-embed: fail on existing defects (clippy collapsible_if in the winit backend, three clippy lints in weft-runtime with wasmtime-runtime, and an E0308 in the selected Servo revision's servo-layout); the runner reports the failing status Add CONTRIBUTING.md, issue forms for defects and changes, maintenance problems and decisions of direction, and a pull request template asking for ownership, proof and the validation actually performed. README.md links the contribution guide.
2026-10-08 23:40:08 +00:00
---
name: weft-integration-verification
description: Use for end-to-end and visual evidence in WEFT OS — real sessions, pixels and native input, failure injection, reliability runs and evidence records.
---
# Integration and visual verification
Integration evidence shows what unit tests cannot: real frames, native input,
real child processes, real components and an installed system. Read the
blueprint's [evidence levels](../../../BLUEPRINT.md#111-evidence-levels),
[verification profiles](../../../BLUEPRINT.md#112-verification-profiles),
[acceptance fixtures](../../../BLUEPRINT.md#113-required-acceptance-fixtures) and
[user journeys](../../../BLUEPRINT.md#22-required-user-journeys).
## Fixture design
- Choose the fixture before implementing. It must fail when the promised
behavior breaks and must not restate the implementation or only grep a success
log.
- For rendering, use a deterministic page with known color regions, text,
scrolling content, a text field, an interactive counter and a controlled
animation. Compare actual pixels read back after painting at the relevant
scales and sizes, including first frame, resize and idle wakeup.
- For input, drive the real native path on a running compositor: key press and
release, modifiers, declared US and Italian layouts with non-ASCII text,
buttons, motion, wheel, focus changes and scale changes. Check for stuck input
after focus loss and client exit.
- For lifecycle, inject failures at each startup phase and after `Running`:
missing binary, invalid package, renderer failure, runtime trap, crash,
cancellation and timeout. Assert a terminal reason and that no owned process,
socket, relay or mount remains.
- For data, use Notes with empty, non-ASCII, multi-line, quoted and literal
backslash text, delayed replies, failed writes and competing writers; repeat
after close, daemon restart, update and reboot.
- Counter and Notes are conformance applications: their happy paths are
necessary but insufficient.
## Running and recording
The [runner](../../scripts/verify.py) provides only build, lint, unit and
package checks. Integration fixtures run in the environment they claim: a nested
winit session for nested-desktop evidence, a DRM session on the declared target,
the booted reference image for installed behavior. Keep captures and logs under
ignored paths such as `target/`; commit fixtures, not their outputs.
Record for each run the commit, environment (host, compositor backend, GPU or
virtual device, Mesa, kernel), enabled features, exact command, observed result
and diagnostics. A screenshot shows pixels, not input routing or accessibility.
An unavailable display, GPU or VM leaves the gate incomplete; say which boundary
remained unverified.
Reliability qualification follows the blueprint targets: repeated clean boots,
launch and close cycles without orphans, mixed interactive and idle runs.
Performance and budget claims need measurements on the reference target.