memory(lfg-acehack-fork): hard-reset safety addendum + content-equivalence audit rule (round-9 peer-harness save)

[AceHack]

Summary

Round-9 multi-AI feedback packet preserved a load-bearing safety rule about AceHack→LFG hard-reset.

What the peer-harness check caught

A prior reset-readiness audit ran git diff --diff-filter=A on origin/main..acehack/main, found no AceHack-only files, and inferred 19/23 PRs were ALREADY-COVERED. The peer-harness check (Codex independently sampling, Grok attacking methodology) caught the gap: file-presence is not content-equivalence. Deeper probe of .github/workflows/gate.yml + .github/codeql/codeql-config.yml found shared files with unique AceHack content NOT on LFG.

Hard-reset is BLOCKED until content-equivalence audit runs cleanly.

Where it lands

Addendum to existing memory/feedback_lfg_master_acehack_zero_divergence_fork_double_hop_aaron_2026_04_27.md (the canonical fork-topology home). No new memory island — consolidation discipline (B-0105) honored.

Key rules (verbatim from the packet)

No AceHack-only files ≠ no AceHack-only content.
Same path is not same substrate.
Hard reset is allowed only after content-loss proof, not content-loss confidence.

Required content-equivalence audit (BEFORE any hard-reset)

# Both checks required:
git diff --name-status --diff-filter=A origin/main..acehack/main   # AceHack-only files
git diff --numstat --diff-filter=M origin/main..acehack/main        # shared files modified
git diff origin/main..acehack/main -- <path>                         # content-compare
git merge-base --is-ancestor <sha> origin/main                       # reachability

Hard-reset gate (binding)

Hard-reset can only return to the table after:
  - all NEEDS-FORWARD-SYNC items have been forwarded to LFG
  - all OBSOLETE / INTENTIONAL-DROP items have rationale recorded
  - all NEEDS-HUMAN-REVIEW items resolved by maintainer sign-off
  - final content-loss audit returns clean

Anti-overcorrection

145 commits is still not the work queue.
Unverified content-loss candidates are the work queue.

Test plan

• Verbatim packet content preserved as addendum (channel rule)
• §33 archive header NOT applicable (in-repo memory file edit, not external-conversation import)
• No new memory file (consolidation discipline)
• MEMORY.md paired-edit updated

🤖 Generated with Claude Code

[Copilot]

Pull request overview

This PR updates the fork-topology “canonical home” memory to include a hard-reset safety addendum clarifying that “no AceHack-only files” is not sufficient evidence of content-equivalence, and records a required audit procedure before any destructive reset.

Changes:

• Added a hard-reset safety addendum describing the prior audit gap and the corrected methodology (including required diff/reachability commands).
• Updated memory/MEMORY.md fast-path note to reflect the paired edit and point readers at the updated canonical memory.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated no comments.

File	Description
memory/feedback_lfg_master_acehack_zero_divergence_fork_double_hop_aaron_2026_04_27.md	Adds the load-bearing “content-equivalence audit” requirement and hard-reset gate criteria as an addendum.
memory/MEMORY.md	Updates the fast-path header comment to reflect the paired edit and keep the index current.