Propose a patch

Evidence

The sandbox observed nothing beyond the static findings; the proposal cites those.

diff --git a/references/onboarding.md b/references/onboarding.md
--- a/references/onboarding.md
+++ b/references/onboarding.md
@@ -132,1 +132,0 @@
-… and the number of historical evidence. 7. After the owner confirms the current write, write `user_memory.md` through the onboarding-only command below. This does not turn on ongoing Draft…
@@ -231,1 +230,0 @@
-5. Clearly tell users: past reply methods and later Draft-edit learning results are saved in the running version `user_memory.md`; view it read-only with:
diff --git a/references/learning-workflow.md b/references/learning-workflow.md
--- a/references/learning-workflow.md
+++ b/references/learning-workflow.md
@@ -122,1 +122,0 @@
-If `memory.usage_enabled=false`, do not add a `user_memory.md` preference or playbook to the Draft-generation context; continue with the standard process.
diff --git a/scripts/configure.py b/scripts/configure.py
--- a/scripts/configure.py
+++ b/scripts/configure.py
@@ -718,1 +718,0 @@
-errors.append("user_memory.md is missing its managed-memory markers")

Zero-context hunks: apply with git apply --unidiff-zero fix.patch on the tagged source of 1.2.8. If the behavior is intended, declare it (host under permissions.network, a line in the README) instead of removing it; the next crawl scores declarations.

Issue / PR body

Title: ecommerce-gmail-customer-service: Instructs the agent to modify its persistent memory or instruction files (KillerSkills verdict caution 60)

## What the honeypot observed
KillerSkills ran **[email protected]** in an isolated sandbox (decoy credentials, recording proxy, syscall trace; engine 1.8.1) and scored it **CAUTION 60/100**. Summary: static analysis only — behavior not yet observed.

- See the findings below.

Full trace and findings: https://killerskills.ai/skills/clawhub/ecommerce-gmail-customer-service/1.2.8

## Findings with a location
- `references/onboarding.md:132` · Instructs the agent to modify its persistent memory or instruction files — references/onboarding.md:132 · Instructs the agent to modify its persistent memory or instruction files · informational: not an instruction file
- `references/learning-workflow.md:122` · Instructs the agent to modify its persistent memory or instruction files — references/learning-workflow.md:122 · Instructs the agent to modify its persistent memory or instruction files · informational: not an instruction file
- `scripts/configure.py:718` · Instructs the agent to modify its persistent memory or instruction files — scripts/configure.py:718 · Instructs the agent to modify its persistent memory or instruction files · informational: not an instruction file
- `references/onboarding.md:231` · Instructs the agent to modify its persistent memory or instruction files — references/onboarding.md:231 · Instructs the agent to modify its persistent memory or instruction files · informational: not an instruction file

## Proposed change
The attached patch removes the 4 lines above. If the behavior is intended, the fix is to declare it instead: list the host under `permissions.network` (or the MCP manifest) and say in the README why the package needs it; the next crawl will re-score it with the declaration taken into account.

## How this will be verified
Nothing is "verified safe" by assertion. When a new version is published, KillerSkills re-runs it in the honeypot and the verdict page updates; you can trigger it early from the verdict page ("Re-scan after I publish a fix"). Reply here or at [email protected] if you believe the verdict is wrong; disputes are reviewed by a person and the outcome is published on the verdict page.

_Proposed through killerskills.ai/threats by a user of the package, under their own GitHub account. The evidence is public; the package was never modified by KillerSkills._