Propose a patch

Evidence

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

diff --git a/src/memcp/templates/vault/patterns.md b/src/memcp/templates/vault/patterns.md
--- a/src/memcp/templates/vault/patterns.md
+++ b/src/memcp/templates/vault/patterns.md
@@ -8,1 +8,0 @@
-The example is indented so the vault parser (which only matches `##` sections at column 0) never reads it. ## AB1 — Short pattern name **Entity:** `Person One` 
diff --git a/src/memcp/templates/vault/entities.md b/src/memcp/templates/vault/entities.md
--- a/src/memcp/templates/vault/entities.md
+++ b/src/memcp/templates/vault/entities.md
@@ -7,1 +7,0 @@
-The example is indented so the vault parser (which only matches `##` cards at column 0) never reads it. ## `Person One` **Role:** their job or relationship to y
diff --git a/src/memcp/templates/vault/timeline.md b/src/memcp/templates/vault/timeline.md
--- a/src/memcp/templates/vault/timeline.md
+++ b/src/memcp/templates/vault/timeline.md
@@ -7,1 +7,0 @@
-The example is indented so the vault parser (which only matches fenced blocks at column 0) never reads it. A fresh scaffold syncs with zero events and zero skip
diff --git a/src/memcp/templates/vault/advice.md b/src/memcp/templates/vault/advice.md
--- a/src/memcp/templates/vault/advice.md
+++ b/src/memcp/templates/vault/advice.md
@@ -7,1 +7,0 @@
-The example is indented so the vault parser (which only matches `##` / `###` sections at column 0) never reads it. ## example-thread ### 2026-01-15 | Short advi
diff --git a/src/memcp/init.py b/src/memcp/init.py
--- a/src/memcp/init.py
+++ b/src/memcp/init.py
@@ -258,1 +258,0 @@
-log.append(f"{action} CLAUDE.md managed block")

Zero-context hunks: apply with git apply --unidiff-zero fix.patch on the tagged source of 0.1.0. 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: memcp: HTML comment carrying imperative instructions to the agent (KillerSkills verdict caution 71)

## 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 71/100**. Summary: no network, touches nothing outside the project, writes no memory.

- See the findings below.

Full trace and findings: https://killerskills.ai/skills/glama/memcp/0.1.0

## Findings with a location
- `src/memcp/templates/vault/patterns.md:8` · HTML comment carrying imperative instructions to the agent — src/memcp/templates/vault/patterns.md:8 · HTML comment carrying imperative instructions to the agent
- `src/memcp/templates/vault/entities.md:7` · HTML comment carrying imperative instructions to the agent — src/memcp/templates/vault/entities.md:7 · HTML comment carrying imperative instructions to the agent
- `src/memcp/templates/vault/timeline.md:7` · HTML comment carrying imperative instructions to the agent — src/memcp/templates/vault/timeline.md:7 · HTML comment carrying imperative instructions to the agent
- `src/memcp/templates/vault/advice.md:7` · HTML comment carrying imperative instructions to the agent — src/memcp/templates/vault/advice.md:7 · HTML comment carrying imperative instructions to the agent
- `src/memcp/init.py:258` · Instructs the agent to modify its persistent memory or instruction files — src/memcp/init.py:258 · Instructs the agent to modify its persistent memory or instruction files · informational: not an instruction file

## Proposed change
The attached patch removes the 5 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._