Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Inline comments:
In `@crates/echo-file-aperture/src/lib.rs`:
- Around line 548-549: The code trusts snapshot.fingerprint.content_digest which
can be inconsistent with snapshot.bytes; recompute the content digest from
snapshot.bytes at the aperture boundary and use that computed value (not
snapshot.fingerprint.content_digest) when deciding posture/verification. In
practice: inside observe and verify_materialization, replace uses of
snapshot.fingerprint.content_digest with a freshly computed digest (e.g., let
computed_digest = compute_content_digest(&snapshot.bytes) or the existing digest
helper used elsewhere), update any local fingerprint/fingerprint-derived
variables from that computed_digest, and base the Unchanged/Verified comparisons
on the recomputed_digest; apply this same change in both observe and
verify_materialization call sites mentioned (the existing variables fingerprint
and posture logic).

In `@crates/echo-file-aperture/tests/file_aperture_tests.rs`:
- Around line 223-244: Add a regression test named
verification_digest_mismatch_returns_materialization_obstruction that constructs
an intentionally inconsistent HostFileSnapshot (mismatched bytes vs fingerprint)
and uses InMemoryFileAperture::observe and verify_materialization to ensure
digests are derived from the snapshot bytes; specifically, create an initial
snapshot via snapshot("/tmp/demo.txt", b"one")?, create a FileContentProposal
using the observed projection (FileContentProposal::new with initial.site_id and
initial.projection.basis), admit it via propose_content, then call
verify_materialization with a snapshot whose bytes differ from the expected
digest (e.g., snapshot("/tmp/demo.txt", b"not-three")?) and assert
MaterializationVerificationPosture::DigestMismatch and that expected_digest !=
observed_digest so the test prevents future reintroduction of digest/bytes
divergence.

In `@docs/BEARING.md`:
- Around line 27-29: Update the "Last updated" header in BEARING.md (the
document header field currently showing "2026-05-25") to the actual date the new
design pointer was added so the signpost metadata is not stale; only change the
header timestamp to the correct updated date and do not alter the new design
pointer entry text.

---

Outside diff comments:
In `@docs/design/echo-owned-file-aperture.md`:
- Around line 514-517: Replace the stale PR status line "Not opened yet for this
branch." (the PR: line in the diff) with the current PR reference and state for
the change (e.g., "Open in PR `#540`" and include the PR URL); ensure the updated
line accurately reflects the open PR number and, optionally, a short status like
"Open" or "Merged" as appropriate and add the PR link for factual
cross-reference.

In `@docs/method/README.md`:
- Around line 128-130: The branch-naming guidance is incomplete: update the
sentence that currently says "Branch names should be lowercase slugs of the
issue title" to also allow explicit cycle work using the cycle/<id>-<slug> form
per METHOD.md and clarify that lowercase slugging still applies to the <slug>
portion; mention the cycle/<id>-<slug> pattern and keep the existing guidance
about lowercase slugs and legend:* labels (or similar) so readers know both
styles are valid.

In `@docs/technical-teardown.md`:
- Around line 1706-1710: The listed command examples use the wrong executable
name `echo-cli`; update each example to use `warp-cli` instead (e.g., change
`echo-cli verify <snapshot> ...` to `warp-cli verify <snapshot> ...`, `echo-cli
inspect` to `warp-cli inspect`, `echo-cli wal doctor` to `warp-cli wal doctor`,
`echo-cli wal submission-posture` to `warp-cli wal submission-posture`, and
`echo-cli bench` to `warp-cli bench`) so the examples match the rest of the
teardown that references `warp-cli`.