Invocation#

bash

spis audit-reference-accessibility --catalog <name> [--catalog <name>]... [--records <selection>] [--batch <id>] [--target <host>] [--plan <path>] [--dry-run]

Plan and run Weles axe accessibility audits over selected captured references.

Inputs and options#

--catalog <name> is required and may be repeated; a call without one is refused as a usage error (audit-reference-accessibility needs at least one --catalog), because no compiled list decides which catalogs are audited. --records <selection> narrows records.
--batch chooses the execution identifier; otherwise Spis generates one.
--target selects a registered Stado host. Without it, Spis uses the active host of the registry's weles-admission service, not a compiled deployment name.
--plan selects a path below .build/accessibility-audit/ in the active corpus. Relative paths resolve against the corpus root; the default is plans/<batch>.json beneath that private work directory.
--dry-run validates and retains only the plan. It contacts no host and is not evidence that an audit ran.

Output and state effects#

Spis writes the validated plan, then stado workload run weles-capture waits for the managed work to finish. Spis reads terminal status, retrieves and verifies evidence, installs the measured record artifacts, invokes verify-reference-evidence --apply --no-state-match, and writes the operational index. Planning, enqueue or terminal-status refusal exits 2; per-reference retrieval or verifier failure exits 3; complete success exits 0.

Private output storage#

Plans live in .build/accessibility-audit/plans/, downloaded staging files in .build/accessibility-audit/staging/<batch>/, and the latest execution index in .build/accessibility-audit/index.json. These paths are relative to the adopted corpus, or the working directory when none is adopted. Exclude .build/ from version control and publication.

The index retains the actual target, plan path, per-record outcome and failed operation. Keep it as private diagnostic evidence. Only validated media and measurements are installed in the selected reference records; a failed operation must not be published as a completed measurement.

bash

spis audit-reference-accessibility --catalog documentation-site-examples --records 1
spis audit-reference-accessibility --catalog documentation-site-examples --records 1 --plan .build/accessibility-audit/plans/review.json

Refusals#

These messages are emitted by the current Rust implementation:

Unknown input is unknown argument: VALUE; every value-taking flag has an exact FLAG needs a value refusal.
An explicit plan outside the private work directory is refused with --plan: PATH is outside ROOT; plans belong under .build/accessibility-audit before dispatch.
If no managed browser host is registered, the diagnostic names the missing weles-admission placement and the supported --target TARGET input.
Unsupported options, including retired polling flags, are usage errors rather than silently ignored inputs.
An empty selection is selection: no records selected.
Plan validation errors exit 2; incomplete/failed retrieval or verifier errors are retained in the index and exit 3.

Real qualification#

The native accessibility-live target uses an actual catalog fixture, a retained state artifact with its measured SHA-256, and Stado/Weles dependencies. It checks private drift-report persistence and a real copied-file hash-mismatch refusal, restores that isolated copy, checks an outside-plan refusal, executes default and relative-plan audits, and compares persisted measurements with downloaded bytes. Source, input and binary hashes, command statuses and evidence remain under ignored .build/accessibility-live-*/. Missing dependencies are failures, not skipped successes. Fixture fields and invocation are documented in the maintainer guide.