Propose a patch

Evidence

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

diff --git a/README.md b/README.md
--- a/README.md
+++ b/README.md
@@ -278,1 +278,0 @@
-<summary><strong>VS Code, Cursor</strong></summary> <br> VS Code reads `.vscode/mcp.json` in the workspace; Cursor reads `~/.cursor/mcp.json` with the same entr
diff --git a/tests/sandbox/test_escapes.py b/tests/sandbox/test_escapes.py
--- a/tests/sandbox/test_escapes.py
+++ b/tests/sandbox/test_escapes.py
@@ -245,1 +245,0 @@
-`~/.netrc` and `~/.git-credentials` are plaintext passwords in a file, and a worker
@@ -249,1 +248,0 @@
-profile's shape, and a test that needed a real `~/.netrc` to exist would be skipped
diff --git a/orchestrator_mcp/ui/view.html b/orchestrator_mcp/ui/view.html
--- a/orchestrator_mcp/ui/view.html
+++ b/orchestrator_mcp/ui/view.html
@@ -2,1 +2,0 @@
-The MCP Apps view for a review or a workflow (ext-apps 2026-01-26). The host renders it in a sandboxed iframe and posts the tool result in; nothing here fetches
diff --git a/tests/test_review_secrets.py b/tests/test_review_secrets.py
--- a/tests/test_review_secrets.py
+++ b/tests/test_review_secrets.py
@@ -114,1 +114,0 @@
-pytest.param("https://example.com:8080/path@x", id="a port is not a password"),

Zero-context hunks: apply with git apply --unidiff-zero fix.patch on the tagged source of 0.12.0. 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: orchestrator-mcp-server: Collapsed <details> block carrying instructions (KillerSkills verdict caution 79)

## 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 79/100**. Summary: no network, touches nothing outside the project, writes no memory.

- See the findings below.

Full trace and findings: https://killerskills.ai/skills/pypi/orchestrator-mcp-server/0.12.0

## Findings with a location
- `README.md:278` · Collapsed <details> block carrying instructions — README.md:278 · Collapsed <details> block carrying instructions
- `tests/sandbox/test_escapes.py:249` · References a credential file in the home directory — tests/sandbox/test_escapes.py:249 · References a credential file in the home directory
- `tests/sandbox/test_escapes.py:245` · References a credential file in the home directory — tests/sandbox/test_escapes.py:245 · References a credential file in the home directory
- `orchestrator_mcp/ui/view.html:2` · HTML comment carrying imperative instructions to the agent — orchestrator_mcp/ui/view.html:2 · HTML comment carrying imperative instructions to the agent · informational: not an instruction file
- `tests/test_review_secrets.py:114` · URL with embedded credentials — tests/test_review_secrets.py:114 · URL with embedded credentials — may be phishing or credential leaking

## 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._