docs: record VT admission rationale only when it passes the record gate #436

Merged
natsukium merged 2 commits from docs/vt-record-rule into main 2026-09-25 12:29:57 +09:00
Owner

docs/explanation/protocols/vt-compliance.md "Versioning the compliance set" required a recorded rationale, in that page, for every added or removed sequence, which conflicts with doc-cascade's "Default to no record" gate.

Per the decision on the issue (option a):

  • The rule now asks for a record only for an admission decision that passes the record gate, and puts it in docs/explanation/protocols/landscape.md, which holds the admission decisions. Any other addition or removal carries its why in the commit body. The support matrix records the status change in the same change.
  • The add-escape-sequence skill's cascade list follows: the compliance reference is updated when a sequence's behavioral facts or caveats change, and landscape.md only under the gate.

Closes #390

`docs/explanation/protocols/vt-compliance.md` "Versioning the compliance set" required a recorded rationale, in that page, for every added or removed sequence, which conflicts with doc-cascade's "Default to no record" gate. Per the decision on the issue (option a): - The rule now asks for a record only for an admission decision that passes the record gate, and puts it in `docs/explanation/protocols/landscape.md`, which holds the admission decisions. Any other addition or removal carries its why in the commit body. The support matrix records the status change in the same change. - The add-escape-sequence skill's cascade list follows: the compliance reference is updated when a sequence's behavioral facts or caveats change, and landscape.md only under the gate. Closes #390
vt-compliance's versioning rule demanded a recorded rationale for every
added or removed sequence, in that page. doc-cascade's "Default to no
record" gate says the opposite for most changes: a why that no
contributor would re-propose against, or that the reference facts
already show, belongs in the commit body. Following the old rule grows
an explanation page with per-sequence notes the gate exists to keep
out, and following the gate breaks the rule.

The rule now asks for a record only for an admission decision that
passes the gate, and puts it in landscape.md, which already holds the
admission decisions and their revisit triggers. Stating the gate as an
explicit exception in doc-cascade was rejected because it keeps a
second record policy alive for one page. The
add-escape-sequence skill's cascade list follows the same rule.

Assisted-by: Claude Code
docs: point the compliance-set rule at the support matrix for status
All checks were successful
darwin / detect relevant changes (pull_request) Successful in 13s
darwin / build felis (aarch64-darwin) (pull_request) Has been skipped
darwin / publish felis (aarch64-darwin) (pull_request) Has been skipped
darwin / darwin docs-only no-op (pull_request) Successful in 0s
windows / detect relevant changes (pull_request) Successful in 1m12s
windows / cargo nextest (Windows) (pull_request) Has been skipped
windows / frontend smoke (Windows) (pull_request) Has been skipped
windows / package felis (x86_64-pc-windows-msvc) (pull_request) Has been skipped
pr / nix flake check (pull_request) Successful in 27s
pr / cargo build / clippy / test / deny (pull_request) Successful in 1m56s
windows / windows docs-only no-op (pull_request) Successful in 2s
pr / wire schema is compatible with the base (pull_request) Successful in 19s
pr / frontend smoke (x86_64-linux) (pull_request) Successful in 1m35s
pr / publish felis (x86_64-linux) (pull_request) Has been skipped
windows / cargo clippy (Windows cross) (pull_request) Successful in 14s
92eefe4795
The support matrix owns per-sequence status; the compliance reference
carries behavior and caveats only. Requiring every added or removed
sequence to be listed in the compliance reference would duplicate
status rows there for sequences with no felis-specific caveat.

Assisted-by: Claude Code
natsukium deleted branch docs/vt-record-rule 2026-09-25 12:29:57 +09:00
Commenting is not possible because the repository is archived.
No description provided.