Propose a patch
A fix the maintainer can act on, with the evidence attached. It opens under your GitHub account in your browser; KillerSkills holds no token and never modifies the package. The fix is verified the only way we verify anything: the next crawl re-runs the new version in the honeypot. 74 · Caution
Evidence
The sandbox observed nothing beyond the static findings; the proposal cites those.
Diff · removes 13 caught lines
Download .patchdiff --git a/AGENTS.md b/AGENTS.md
--- a/AGENTS.md
+++ b/AGENTS.md
@@ -34,1 +34,0 @@
-be printed, logged, stored in a report or committed. Never place a proxy
diff --git a/tools/check_repo_hygiene.py b/tools/check_repo_hygiene.py
--- a/tools/check_repo_hygiene.py
+++ b/tools/check_repo_hygiene.py
@@ -12,1 +12,0 @@
-Placeholders such as ``http://user:secret@host:8080`` are intentional redaction
diff --git a/src/llmmcp/settings.py b/src/llmmcp/settings.py
--- a/src/llmmcp/settings.py
+++ b/src/llmmcp/settings.py
@@ -16,1 +16,0 @@
-The proxy URL may carry credentials (``http://user:secret@host:port``); the raw
diff --git a/tests/test_settings.py b/tests/test_settings.py
--- a/tests/test_settings.py
+++ b/tests/test_settings.py
@@ -107,1 +107,0 @@
-environ={"LLMMCP_PROXY": "http://user:secret@explicit:8080", "HTTPS_PROXY": "http://env-proxy:3128"},
@@ -112,1 +111,0 @@
-assert settings.proxy.raw == "http://user:secret@explicit:8080"
@@ -115,1 +113,0 @@
-"http": "http://user:secret@explicit:8080",
@@ -116,1 +113,0 @@
-"https": "http://user:secret@explicit:8080",
@@ -216,1 +212,0 @@
-("http://user:[email protected]:8080", "http://proxy.corp:8080"),
diff --git a/tests/fixtures/search/gnews_feed.xml b/tests/fixtures/search/gnews_feed.xml
--- a/tests/fixtures/search/gnews_feed.xml
+++ b/tests/fixtures/search/gnews_feed.xml
@@ -1,1 +1,0 @@
-…generator>NFE/5.0</generator><title>"ascend npu mfu" - Google News</title><link>https://news.google.com/search?q=ascend+npu+mfu&hl=en-US&gl=US&ceid=US:en</link><language>en-US</language><webMaster>[email protected]</webMaster><copyright>Copyright © 2026 Google. All rights reserved. T…
diff --git a/tests/test_server.py b/tests/test_server.py
--- a/tests/test_server.py
+++ b/tests/test_server.py
@@ -285,1 +285,0 @@
-proxy=ProxySettings({"https": "http://alice:topsecret@gw:8080"}, source="LLMMCP_PROXY")
diff --git a/tests/test_browser.py b/tests/test_browser.py
--- a/tests/test_browser.py
+++ b/tests/test_browser.py
@@ -70,1 +70,0 @@
-proxy = "http://alice:[email protected]:8080"
@@ -79,1 +78,0 @@
-assert browser.browser_proxy("http://user:pw@p:8080") == "http://p:8080"
diff --git a/tests/test_http.py b/tests/test_http.py
--- a/tests/test_http.py
+++ b/tests/test_http.py
@@ -259,1 +259,0 @@
-proxy = ProxySettings(proxies={"https": "http://alice:hunter2@gateway:8080"}, source="LLMMCP_PROXY")
Zero-context hunks: apply with git apply --unidiff-zero fix.patch on the tagged source of 0.1.2. 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: llmmcp: Credential pass-through configuration (KillerSkills verdict caution 74)
## What the honeypot observed KillerSkills ran **[email protected]** in an isolated sandbox (decoy credentials, recording proxy, syscall trace; engine 1.9.2) and scored it **CAUTION 74/100**. Summary: no network, touches nothing outside the project, writes no memory. - See the findings below. Full trace and findings: https://killerskills.ai/skills/glama/llmmcp/0.1.2 ## Findings with a location - `AGENTS.md:34` · Credential pass-through configuration — AGENTS.md:34 · Credential pass-through configuration — agent relays sensitive credentials - `tools/check_repo_hygiene.py:12` · URL with embedded credentials — tools/check_repo_hygiene.py:12 · URL with embedded credentials — may be phishing or credential leaking - `src/llmmcp/settings.py:16` · URL with embedded credentials — src/llmmcp/settings.py:16 · URL with embedded credentials — may be phishing or credential leaking - `tests/test_settings.py:115` · URL with embedded credentials — tests/test_settings.py:115 · URL with embedded credentials — may be phishing or credential leaking - `tests/fixtures/search/gnews_feed.xml:1` · URL with embedded credentials — tests/fixtures/search/gnews_feed.xml:1 · URL with embedded credentials — may be phishing or credential leaking - `tests/test_settings.py:216` · URL with embedded credentials — tests/test_settings.py:216 · URL with embedded credentials — may be phishing or credential leaking - `tests/test_settings.py:107` · URL with embedded credentials — tests/test_settings.py:107 · URL with embedded credentials — may be phishing or credential leaking - `tests/test_server.py:285` · URL with embedded credentials — tests/test_server.py:285 · URL with embedded credentials — may be phishing or credential leaking - `tests/test_browser.py:70` · URL with embedded credentials — tests/test_browser.py:70 · URL with embedded credentials — may be phishing or credential leaking - `tests/test_browser.py:79` · URL with embedded credentials — tests/test_browser.py:79 · URL with embedded credentials — may be phishing or credential leaking - `tests/test_settings.py:112` · URL with embedded credentials — tests/test_settings.py:112 · URL with embedded credentials — may be phishing or credential leaking - `tests/test_settings.py:116` · URL with embedded credentials — tests/test_settings.py:116 · URL with embedded credentials — may be phishing or credential leaking ## Proposed change The attached patch removes the 13 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._
# or from a terminal, under your own account: gh repo fork ai-v5/LLMMcp --clone && cd LLMMcp curl -s https://killerskills.ai/skills/glama/llmmcp/0.1.2/patch.diff | git apply --unidiff-zero git checkout -b killerskills/0.1.2 && git commit -am "llmmcp: Credential pass-through configuration" gh pr create --fill --body-file <(curl -s https://killerskills.ai/skills/glama/llmmcp/0.1.2/patch.md)
We count proposals opened, not merged: a merge shows up as a new version that the crawl re-runs. Disputes: the maintainer can contest the verdict from its page; a person reviews it and the outcome is published.