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
@@ -110,1 +110,0 @@
-can write the summaries ([Agents](docs/agents.md)). To connect the AI apps on this computer (Claude Code,
@@ -133,1 +132,0 @@
-<summary><b>Other ways to run it: from source, Docker</b></summary> From source, to hack on it or run it as a login service: ```bash git clone https://github.co
diff --git a/packages/core/src/import.ts b/packages/core/src/import.ts
--- a/packages/core/src/import.ts
+++ b/packages/core/src/import.ts
@@ -116,1 +116,0 @@
-const s = src.replace(/^/, "");
@@ -198,1 +197,0 @@
-const head = src.replace(/^/, "").trimStart();
diff --git a/packages/core/src/guard.ts b/packages/core/src/guard.ts
--- a/packages/core/src/guard.ts
+++ b/packages/core/src/guard.ts
@@ -15,1 +15,0 @@
-* U+200D (zero-width joiner) is kept: emoji such as 🙅🏻‍♀️ are built with it.
diff --git a/scripts/phone-shot.html b/scripts/phone-shot.html
--- a/scripts/phone-shot.html
+++ b/scripts/phone-shot.html
@@ -1,1 +1,0 @@
-Phone triptych for the README. Run apps/server/e2e/web.mjs with SCREENSHOTS=<dir>, put this file in that dir, replace LANG and CAP1-3, screenshot at 1240x900 @2
diff --git a/scripts/extension-shot.html b/scripts/extension-shot.html
--- a/scripts/extension-shot.html
+++ b/scripts/extension-shot.html
@@ -1,1 +1,0 @@
-Composes the README extension screenshot. Run apps/chrome-extension/e2e.mjs with SCREENSHOTS=<dir> (and SHOT_LANG), replace SHOTS and LANG below, then screensho

Zero-context hunks: apply with git apply --unidiff-zero fix.patch on the tagged source of r1. 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: driftnet: Collapsed <details> block carrying instructions (KillerSkills verdict caution 63)

## What the honeypot observed
KillerSkills ran **driftnet@r1** in an isolated sandbox (decoy credentials, recording proxy, syscall trace; engine 1.8.1) and scored it **CAUTION 63/100**. Summary: no network, touches nothing outside the project, writes no memory.

- See the findings below.

Full trace and findings: https://killerskills.ai/skills/glama/driftnet/r1

## Findings with a location
- `README.md:133` · Collapsed <details> block carrying instructions — README.md:133 · Collapsed <details> block carrying instructions
- `packages/core/src/import.ts:198` · Hidden zero-width characters detected — packages/core/src/import.ts:198 · Hidden zero-width characters detected — may conceal content
- `packages/core/src/guard.ts:15` · Hidden zero-width characters detected — packages/core/src/guard.ts:15 · Hidden zero-width characters detected — may conceal content
- `packages/core/src/import.ts:116` · Hidden zero-width characters detected — packages/core/src/import.ts:116 · Hidden zero-width characters detected — may conceal content
- `scripts/phone-shot.html:1` · HTML comment carrying imperative instructions to the agent — scripts/phone-shot.html:1 · HTML comment carrying imperative instructions to the agent · informational: not an instruction file
- `README.md:110` · Instructs the agent to modify its persistent memory or instruction files — README.md:110 · Instructs the agent to modify its persistent memory or instruction files · informational: not an instruction file
- `scripts/extension-shot.html:1` · HTML comment carrying imperative instructions to the agent — scripts/extension-shot.html:1 · HTML comment carrying imperative instructions to the agent · informational: not an instruction file

## Proposed change
The attached patch removes the 7 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._