gates: duplicate-bodies, scan and t27b-native-ratchet are red on master (has_at, hw_place_before, a home path, 10 unlisted specs) #8361

Three gates are red on master's head (8881377) and therefore on every open pull request. Each one has a single cause in a recent merge.

Goal

Make duplicate-bodies, scan and t27b-native-ratchet pass on master again, each by the narrowest change that removes its own cause, so that pull requests are judged on their own diff.

Where we are

• duplicate-bodies: python3 tools/dupe_scan.py on master prints a body written under has_at is now copied [2] time(s) (t27b: the body walks of lower.rs move to specs/tri/t27b/ast_walk.t27 (Closes #8249) #8258: specs/tri/t27b/ast_walk.t27 has_at = specs/tri/t27b/source_text.t27 starts_with) and hw_place_before ... [2] (d2d1112: specs/queen/actors_hw.t27 hw_place_before = specs/queen/actors.t27 place_before, at u16 instead of u64 node ids).
• scan: tools/check_devhome_paths.py fails on .claude/skills/t27b-loop/SKILL.md (skills: t27b-loop, one tick of the unattended t27b improvement loop (Closes #8331) #8332), which names /Users/playom/t27.
• t27b-native-ratchet: 10 specs are UNLISTED in docs/reports/t27b_expectations.json (for example specs/queen/keyed.t27, specs/math/constants.t27), seen on PR runs such as 37988934908.

Boundary

• Changed: tools/duplicate_bodies_baseline.txt, .claude/skills/t27b-loop/SKILL.md, docs/reports/t27b_expectations.json.
• Out of bounds: every spec, the compiler, other ledgers.

User Scenarios & Testing

1. A pull request that touches none of these files gets green duplicate-bodies, scan and t27b-native-ratchet on its merge commit.
2. A new copied body, a new home path or a new spec without a ledger row still fails its gate.

Requirements

• R1. has_at cannot be replaced by an import: t27c gen-rust emits the call to starts_with without a use, and cli/t27b/src/lower.rs mounts ast_walk.rs as its own module, so cli/t27b would not compile. hw_place_before is the hardware-width profile of place_before (actors_hw.t27 exists for gen-verilog). Both are recorded as deliberate copies; no other group moves.
• R2. The skill reads git config --get core.hooksPath, which every worktree shares, instead of a developer's home path.
• R3. The t27b rows come from a t27b-native run of this branch's own head, written with tri t27b ratchet --bless --accept-new, and only rows the reference passes are recorded.

Success Criteria

• python3 tools/dupe_scan.py and python3 tools/check_devhome_paths.py pass on the branch, and the baseline diff is exactly two added lines.
• t27b-native-ratchet is green on the branch head.
• Required checks green.

Requirements

FR-001: The duplicate-bodies baseline MUST record specs/tri/t27b/ast_walk.t27 has_at = specs/tri/t27b/source_text.t27 starts_with and specs/queen/actors_hw.t27 hw_place_before = specs/queen/actors.t27 place_before as deliberate copies; no import replacement is permitted because t27c gen-rust emits the call to starts_with without a use and cli/t27b/src/lower.rs mounts ast_walk.rs as its own module.

FR-002: The skill file .claude/skills/t27b-loop/SKILL.md MUST read git config --get core.hooksPath instead of a developer's home path.

FR-003: The ledger docs/reports/t27b_expectations.json MUST be updated by running tri t27b ratchet --bless --accept-new on the branch's own head, recording only rows the reference passes.