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
@@ -34,1 +34,0 @@
-Tool tests run on two local models, three times each, with a failed run's wrong argument
diff --git a/dist/web/assets/html-BcACZWSP.js b/dist/web/assets/html-BcACZWSP.js
--- a/dist/web/assets/html-BcACZWSP.js
+++ b/dist/web/assets/html-BcACZWSP.js
@@ -1,1 +1,0 @@
-…ude":"#math"},{"include":"#svg"},{"begin":"(<)([A-Za-z][.0-9A-Z_a-z·À-ÖØ-öø-ͽͿ-῿‌‍‿⁀⁰-↏Ⰰ-⿯、-퟿豈-﷏ﷰ-�𐀀-\\\\x{EFFFF}]*-[-.0-9A-Z_a-z·À-ÖØ-öø-ͽͿ-῿‌‍‿⁀⁰-↏Ⰰ-⿯、-퟿豈-﷏ﷰ-…
diff --git a/dist/web/assets/core-C0nYwqGS.js b/dist/web/assets/core-C0nYwqGS.js
--- a/dist/web/assets/core-C0nYwqGS.js
+++ b/dist/web/assets/core-C0nYwqGS.js
@@ -3,1 +3,0 @@
-…¤`,yen:`¥`,brvbar:`¦`,sect:`§`,uml:`¨`,copy:`©`,ordf:`ª`,laquo:`«`,not:`¬`,shy:`­`,reg:`®`,macr:`¯`,deg:`°`,plusmn:`±`,sup2:`²`,sup3:`³`,acute:`´`,micro:`µ`,para…
diff --git a/dist/web/assets/katex-HrBcstkf.js b/dist/web/assets/katex-HrBcstkf.js
--- a/dist/web/assets/katex-HrBcstkf.js
+++ b/dist/web/assets/katex-HrBcstkf.js
@@ -251,1 +251,0 @@
-…t`],[G([],[])]);x.style.height=A(-u),b=[G([`vlist-r`],[y,G([`vlist-s`],[new Le(`​`)])]),G([`vlist-r`],[x])]}else b=[G([`vlist-r`],[y])];var S=G([`vlist-t`],b);re…
diff --git a/dist/web/assets/less-DVTAwKKz.js b/dist/web/assets/less-DVTAwKKz.js
--- a/dist/web/assets/less-DVTAwKKz.js
+++ b/dist/web/assets/less-DVTAwKKz.js
@@ -1,1 +1,0 @@
-…lation"}]},{"include":"#style-function"},{"match":"--|-?(?:[A-Z_a-z·À-ÖØ-öø-ͽͿ-῿‌‍‿⁀⁰-↏Ⰰ-⿯、-퟿豈-﷏ﷰ-�𐀀-\\\\x{EFFFF}]|\\\\\\\\(?:\\\\N|\\\\H|\\\\h{1,6}[R\\\\s]))(?…

Zero-context hunks: apply with git apply --unidiff-zero fix.patch on the tagged source of 0.6.0. 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: sandbox: Image alt text carrying instructions (KillerSkills verdict caution 62)

## 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 62/100**. Summary: no network, touches nothing outside the project, writes no memory.

- See the findings below.

Full trace and findings: https://killerskills.ai/skills/npm/sandbox/0.6.0

## Findings with a location
- `README.md:34` · Image alt text carrying instructions — README.md:34 · Image alt text carrying instructions
- `dist/web/assets/html-BcACZWSP.js:1` · Hidden zero-width characters detected — dist/web/assets/html-BcACZWSP.js:1 · Hidden zero-width characters detected — may conceal content
- `dist/web/assets/core-C0nYwqGS.js:3` · Hidden zero-width characters detected — dist/web/assets/core-C0nYwqGS.js:3 · Hidden zero-width characters detected — may conceal content
- `dist/web/assets/katex-HrBcstkf.js:251` · Hidden zero-width characters detected — dist/web/assets/katex-HrBcstkf.js:251 · Hidden zero-width characters detected — may conceal content
- `dist/web/assets/pdf-Ci7i7T5q.js:3` · Base64 encoding combined with network call — dist/web/assets/pdf-Ci7i7T5q.js:3 · Base64 encoding combined with network call — potential data exfiltration
- `dist/web/assets/less-DVTAwKKz.js:1` · Hidden zero-width characters detected — dist/web/assets/less-DVTAwKKz.js:1 · 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._