[chaksaray]

Closes #74

What and why

Maintainer decision in #71: the gate targets AVE schema 1.1.0. DESIGN.md 9 specified an AVE 1.2 that AVE never published.

Measured, aveproject/ave main and gate origin/develop, 2026-10-18/19:

dist/ave-records-latest.json      87 records, schema_version 1.1.0 for all, AIVSS spec 0.8
provenance_vector                 87/87; entry_class is a layer token (tool_response, server_card, ...), escalation data_to_instruction style
trifecta_profile                  77/87; requires + amplifies arrays
capability_mitigation             not in schema/ave-record-1.1.0.schema.json; 0/87 records
harden --ave                      reads mitigations/ave-mitigations.json (2 entries: manifest_stanza, review_status, reviewed_by, reviewed_at, rule_ids)
src/ reads provenance_vector / trifecta_profile   no (git grep: only tests/, scripts/)

Changes, DESIGN.md only:

• Header: "Depends on: AVE schema 1.1.0 (aveproject/ave; see Section 9)"
• 6.2: data.ave cites IDs whose mitigation stanza produced the blocking rule (Section 9.1)
• 9: "AVE records: schema 1.1.0". The gate consumes AVE 1.1.0 and does not define or extend the AVE schema
• 9.1: the fields the gate uses, by reference to AVE's schema; mitigation stanzas live in the gate's own mitigations map, with the shape of schemas/bawbel-ave-mitigations-v1.json; the gate does not consume provenance_vector or trifecta_profile at runtime
• 9.2: marked superseded, with what happened. This also removes the last ave.bawbel.io string in the repo
• 9.3: unchanged
• 12, Open: how AVE entry_class tokens map to gate provenance classes. Not specified here, on purpose: that mapping is a design choice, not a fact

Not changed: Section 11 still cites BAWBEL_GATE_MITIGATIONS_SPEC.md Section 3, which is gitignored (.gitignore:2) and so not in the repository.

Governing spec

DESIGN.md section(s): header, 6.2, 9 (rewritten), 9.2 (superseded), 12.

Output contract

• Changes: DESIGN.md
• Does not change: code, tests, schemas (that is Move the code to AVE schema 1.1.0 (after the DESIGN.md 9 change) #75)
• New or changed CLI flags, manifest fields, reason strings, audit record fields: none

Verification

$ grep -n -E 'AVE schema >= 1\.2|ave\.bawbel\.io' DESIGN.md
(no output)

pytest                          -> 346 passed in 11.60s
pytest tests/property/ -x       -> 26 passed in 1.42s
pytest tests/replay/            -> 32 passed in 0.35s
pytest tests/test_m4_corpus.py  -> 10 passed in 0.17s   (approval budget)
pre-commit run --files DESIGN.md -> Passed
em dashes on added lines: 0

Checklist

• Every commit carries a DCO sign-off (git commit -s)
• DESIGN.md is still accurate after this change (code catches up in Move the code to AVE schema 1.1.0 (after the DESIGN.md 9 change) #75)
• No em dashes in docs, comments, or error strings

🤖 Generated with Claude Code