Propose a patch

Evidence

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

diff --git a/dist/src/dist2.js b/dist/src/dist2.js
--- a/dist/src/dist2.js
+++ b/dist/src/dist2.js
@@ -1,1 +1,0 @@
-…:`≠`,nearhk:`⤤`,neArr:`⇗`,nearr:`↗`,nearrow:`↗`,nedot:`≐̸`,NegativeMediumSpace:`​`,NegativeThickSpace:`​`,NegativeThinSpace:`​`,NegativeVeryThinSpace:`​`,nequiv:…
@@ -2,1 +1,0 @@
-…ow:`←`,shortmid:`∣`,shortparallel:`∥`,ShortRightArrow:`→`,ShortUpArrow:`↑`,shy:`­`,Sigma:`Σ`,sigma:`σ`,sigmaf:`ς`,sigmav:`ς`,sim:`∼`,simdot:`⩪`,sime:`≃`,simeq:`≃…
diff --git a/dist/src/undici.js b/dist/src/undici.js
--- a/dist/src/undici.js
+++ b/dist/src/undici.js
@@ -12,1 +12,0 @@
-…for(let e of r)t.push(Buffer.from(n),Buffer.from(e));else t.push(Buffer.from(n),Buffer.from(r));return t}function Z(e,t){e[E]=t;let n=e[M],r=e[N],i=ne.connect(e[_],{createConnection:()=>t,peerMaxConcurrentStreams:e[k],settings:{enablePush…
diff --git a/dist/src/proxy.js b/dist/src/proxy.js
--- a/dist/src/proxy.js
+++ b/dist/src/proxy.js
@@ -1,1 +1,0 @@
-…===`/rpc`?Response.json({jsonrpc:`2.0`,id:r,error:{code:-32001,message:`Invalid proxy token`,data:t(new e(`UNAUTHORIZED`,`Invalid proxy token`))}},{status:401}):Response.j…
diff --git a/README.md b/README.md
--- a/README.md
+++ b/README.md
@@ -18,1 +18,0 @@
-Intro rule: category line, job pitch with the platform list, works-with/proof lines, nothing else. Per-platform transports, tool names, vendor lists, and suppor

Zero-context hunks: apply with git apply --unidiff-zero fix.patch on the tagged source of 0.21.37. 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: agent-device: Hidden zero-width characters detected (KillerSkills verdict caution 64)

## 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 64/100**. Summary: no network, touches nothing outside the project, writes no memory.

- See the findings below.

Full trace and findings: https://killerskills.ai/skills/npm/agent-device/0.21.37

## Findings with a location
- `dist/src/dist2.js:1` · Hidden zero-width characters detected — dist/src/dist2.js:1 · Hidden zero-width characters detected — may conceal content
- `dist/src/undici.js:12` · Host or code assembled from base64 fragments — dist/src/undici.js:12 · Host or code assembled from base64 fragments
- `dist/src/proxy.js:1` · Credential pass-through configuration — dist/src/proxy.js:1 · Credential pass-through configuration — agent relays sensitive credentials
- `README.md:18` · HTML comment carrying imperative instructions to the agent — README.md:18 · HTML comment carrying imperative instructions to the agent
- `dist/src/dist2.js:2` · Hidden zero-width characters detected — dist/src/dist2.js:2 · Hidden zero-width characters detected — may conceal content

## 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._