fix(memory): skip drift guard for add (append-only) action

[liuhao1024]

What does this PR do?

Skips the drift guard for memory(action=add) which only appends and never overwrites existing content. The drift guard (introduced for #26045) correctly protects replace/remove from clobbering un-roundtrippable content, but firing it on add causes false positives when prior add() calls in the same session shift the byte count of the on-disk file.

Related Issue

Fixes #42874

Type of Change

• 🐛 Bug fix (non-breaking change that fixes an issue)

Changes Made

• tools/memory_tool.py: Added skip_drift keyword parameter to _reload_target(); add() passes skip_drift=True while replace()/remove() continue using the drift guard unchanged
• tests/tools/test_memory_tool.py: Updated test_add_succeeds_despite_drift (was test_add_refuses_on_drift) to verify add succeeds under drift; updated test_drift_backup_filename_is_unique_per_invocation to use replace/remove since add no longer triggers drift

How to Test

1. Start a fresh Hermes session with existing MEMORY.md content
2. Perform multiple memory(action=add) calls in the same session
3. Verify subsequent add calls succeed without "Refusing to write" errors
4. Verify memory(action=replace) still refuses when external drift is present
5. Run pytest tests/tools/test_memory_tool.py -v — all 69 tests pass

Checklist

Code

• I've read the Contributing Guide
• My commit messages follow Conventional Commits (fix(scope):, feat(scope):, etc.)
• I searched for existing PRs to make sure this isn't a duplicate
• My PR contains only changes related to this fix/feature (no unrelated commits)
• I've run pytest tests/tools/test_memory_tool.py -q and all tests pass
• I've added tests for my changes (required for bug fixes, strongly encouraged for features)
• I've tested on my platform: macOS

Documentation & Housekeeping

• I've updated relevant documentation (README, docs/, docstrings) — or N/A
• I've updated cli-config.yaml.example if I added/changed config keys — or N/A
• I've updated CONTRIBUTING.md or AGENTS.md if I changed architecture or workflows — or N/A
• I've considered cross-platform impact (Windows, macOS) per the compatibility guide — or N/A
• I've updated tool descriptions/schemas if I changed tool behavior — or N/A

Code Intelligence

• Analyzed: tools/memory_tool.py — _reload_target() (callers: 3), _detect_external_drift() (callers: 1)
• Blast radius: LOW — only add() path changed; replace()/remove() unchanged
• Related patterns: Drift guard (memory(action=replace) silently clobbers external writes to MEMORY.md (data-loss race with patch tool / shell appends / concurrent sessions) #26045), append-only vs full-rewrite semantics

[tonydwb]

Code Review Summary

Verdict: Approved

Changes Overview

• Fixes issue #42874: add (append-only) action no longer triggers drift detection
• Adds skip_drift parameter to _reload_target method
• Updates tests to reflect the corrected behavior

Analysis

Correctness

• The fix is logically sound: drift detection guards against clobbering un-roundtrippable content during replace/remove (which rewrite the full file), but add only appends — it can never overwrite existing content
• The skip_drift=True parameter is cleanly implemented and well-documented in the docstring

Security

• No security concerns: this is a pure in-memory/file storage fix with no user-input or network components

Code Quality

• Well-scoped single concern (issue #42874)
• Clear documentation explains the rationale for the behavior change
• Good test coverage with descriptive test name (test_add_succeeds_despite_drift)
• Test comment clarifies the intended behavior vs. the old (incorrect) behavior

Testing

• Dedicated test for the new behavior
• Existing test (test_drift_backup_filename_is_unique_per_invocation) updated to use remove instead of add for the second drift-inducing operation (since add no longer triggers drift)

Minor Note

• The existing test name test_drift_backup_filename_is_unique_per_invocation still validates that unique .bak filenames are created, but now tests the remove path instead of add. The comment added to that test helps clarify the intent.

---

Reviewed by Hermes Agent

[teknium1]

Automated hermes-sweeper review: this PR's fix has already been implemented on main.

Evidence:

• tools/memory_tool.py:251 now has _reload_target(..., skip_drift=False) and bypasses drift detection when skip_drift=True.
• tools/memory_tool.py:311 shows MemoryStore.add() calling _reload_target(target, skip_drift=True), while replace/remove still use the drift guard at tools/memory_tool.py:366 and tools/memory_tool.py:423.
• tests/tools/test_memory_tool.py:607 includes test_add_succeeds_despite_drift, covering the behavior this PR requested.
• Commit 25e2312230ca95f843ee31a93c65899a9fe01272 landed the fix and explicitly notes it was salvaged from fix(memory): skip drift guard for add (append-only) action #42880 by @liuhao1024.

Thanks for the original fix — it made it into main.