[marcuspat]

Third CI follow-up to #11 (rustc broken locally, CI is the compiler here): clippy denies a redundant closure (|p| normalise_components(p) → pass the fn) and the now-useless .map(String::from). No semantic change.

[marcuspat]

Gate review — reviewer: claude

Lanes: reviewed by claude

GATE USAGE reviewer=claude in=2 cached=41,795 out=909 14s model=claude-haiku-4-5-20251001,claude-opus-5[1m] $0.322 API-equiv (not billed on subscription)

Pure clippy/rustfmt cleanup. Walking the two semantic changes:

sentinel-capabilities/src/filesystem.rs:265 — failure("...", false) replacing failure("...".into(), false). The PR body (base commit 88115df) says failure takes impl Into<String>, so a &str literal coerces identically. No behavior change.

sentinel-capabilities/src/filesystem.rs:316 — the interesting one. .filter_map(|p| normalise_components(p)).filter(...).map(String::from) → .filter_map(normalise_components).filter(...). Dropping .map(String::from) is only correct if normalise_components already returns Option<String> rather than Option<&str>. Two pieces of evidence in the diff say it does: line 264 binds let Some(log_dir_norm) = normalise_components(log_dir) then let log_dir: &str = &log_dir_norm; — that reborrow is only needed if the value is owned; and line 564 calls .unwrap_or_default() on the result, which requires the inner type to be Default (&str is, via &"", so that one's weaker, but combined with the reborrow it's conclusive). So the old .map(String::from) was a String→String no-op — exactly the clippy::useless_conversion/redundant-clone class this PR is clearing. The closure-to-path change (redundant_closure) is sound because normalise_components takes a single arg by value/ref matching the iterator item; if the types didn't line up it wouldn't compile, and the collect target is unchanged.

The prefix filter semantics are untouched: starts_with(&prefix) still runs on the normalised path, so the traversal guard the comment at :262-264 describes is intact.

sentinel-exec/src/sandbox.rs — all four hunks are rustfmt rewrapping of a for loop array literal and three assert macro calls. Message strings, assertion operands, and the t(i, jt) index arithmetic are byte-identical across the rewrap. The jumps >= 5 guard (the "this test verified nothing" check) and the six assert_eq! target checks keep their original values.

No test changes needed — this diff changes no behavior, and the one place it touches test code (filesystem.rs:561-566, sandbox.rs) preserves every assertion's operands and messages.

One note, not a finding: I can't verify from the diff that cargo clippy -D warnings now passes, only that these edits target the usual lint classes. The commit message claims CI is fixed; that's unverified here.

VERDICT: APPROVED

---

VERDICT: APPROVED