Propose a patch

Evidence

• Read ~/.ssh-manager/profile in the decoy home directory.

diff --git a/src/server-import.js b/src/server-import.js
--- a/src/server-import.js
+++ b/src/server-import.js
@@ -176,1 +176,0 @@
-label: 'OpenSSH config (~/.ssh/config)',
diff --git a/README.md b/README.md
--- a/README.md
+++ b/README.md
@@ -425,1 +425,0 @@
-<summary><b>Spending less context — tool groups</b></summary> <br> All 37 tools cost roughly **43.5k tokens** of context. Most people don't need all of them. ``
diff --git a/src/tool-config-manager.js b/src/tool-config-manager.js
--- a/src/tool-config-manager.js
+++ b/src/tool-config-manager.js
@@ -16,1 +16,0 @@
-* User-global only: ~/.ssh-manager/tools-config.json
diff --git a/src/server-export.js b/src/server-export.js
--- a/src/server-export.js
+++ b/src/server-export.js
@@ -87,1 +87,0 @@
-keyPath: '~/.ssh/id_ed25519', defaultDir: '/var/www', group: 'production',
@@ -90,1 +89,0 @@
-keyPath: '~/.ssh/id_ed25519', group: 'staging' },
diff --git a/dist/ui/index.html b/dist/ui/index.html
--- a/dist/ui/index.html
+++ b/dist/ui/index.html
@@ -6,1 +6,0 @@
-Inlined: the browser requests a favicon without the token, which the control plane refuses. A data URI costs no request at all.

Zero-context hunks: apply with git apply --unidiff-zero fix.patch on the tagged source of 4.0.3. 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: mcp-ssh-manager: Read ~/.ssh-manager/profile in the decoy home directory (KillerSkills verdict blocked 37)

## 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 **BLOCKED 37/100**. Summary: reads credential files.

- Read ~/.ssh-manager/profile in the decoy home directory.

Full trace and findings: https://killerskills.ai/skills/npm/mcp-ssh-manager/4.0.3

## Findings with a location
- `src/server-import.js:176` · Reads credential paths from the home directory — src/server-import.js:176 · Reads credential paths from the home directory
- `README.md:425` · Collapsed <details> block carrying instructions — README.md:425 · Collapsed <details> block carrying instructions
- `src/tool-config-manager.js:16` · Reads credential paths from the home directory — src/tool-config-manager.js:16 · Reads credential paths from the home directory
- `src/server-export.js:90` · References a credential file in the home directory — src/server-export.js:90 · References a credential file in the home directory
- `src/server-export.js:87` · References a credential file in the home directory — src/server-export.js:87 · References a credential file in the home directory
- `dist/ui/index.html:6` · HTML comment carrying imperative instructions to the agent — dist/ui/index.html:6 · HTML comment carrying imperative instructions to the agent · informational: not an instruction file

## Proposed change
The attached patch removes the 6 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._