Propose a patch

Evidence

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

diff --git a/rules/AGENTS.md b/rules/AGENTS.md
--- a/rules/AGENTS.md
+++ b/rules/AGENTS.md
@@ -1,1 +1,0 @@
-Rule: Token-Optimized Communication Protocol (TOCP) Trigger: always_on
diff --git a/skills/token-optimizer/references/cli-reference.md b/skills/token-optimizer/references/cli-reference.md
--- a/skills/token-optimizer/references/cli-reference.md
+++ b/skills/token-optimizer/references/cli-reference.md
@@ -30,1 +30,0 @@
-tokensafe test "<prompt>"     # Test TOCP S-expression prompt compaction
diff --git a/README.md b/README.md
--- a/README.md
+++ b/README.md
@@ -374,1 +374,0 @@
-tokensafe test "<prompt>"           Preview prompt compaction on custom input

Zero-context hunks: apply with git apply --unidiff-zero fix.patch on the tagged source of 3.1.5. 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: tokensafe: HTML comment carrying imperative instructions to the agent (KillerSkills verdict caution 52)

## 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 52/100**. Summary: no network, touches nothing outside the project, writes no memory.

- See the findings below.

Full trace and findings: https://killerskills.ai/skills/npm/tokensafe/3.1.5

## Findings with a location
- `rules/AGENTS.md:1` · HTML comment carrying imperative instructions to the agent — rules/AGENTS.md:1 · HTML comment carrying imperative instructions to the agent
- `skills/token-optimizer/references/cli-reference.md:30` · Fake instruction boundary tag — skills/token-optimizer/references/cli-reference.md:30 · Fake instruction boundary tag · informational: not an instruction file
- `README.md:374` · Fake instruction boundary tag — README.md:374 · Fake instruction boundary tag · informational: not an instruction file

## Proposed change
The attached patch removes the 3 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._