Propose a patch

Evidence

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

diff --git a/.claude/skills/openspec-propose/SKILL.md b/.claude/skills/openspec-propose/SKILL.md
--- a/.claude/skills/openspec-propose/SKILL.md
+++ b/.claude/skills/openspec-propose/SKILL.md
@@ -166,1 +166,0 @@
-- Do NOT copy `<context>`, `<rules>`, `<project_context>` blocks into the artifact
diff --git a/packages/screen/README.md b/packages/screen/README.md
--- a/packages/screen/README.md
+++ b/packages/screen/README.md
@@ -3,1 +3,0 @@
-Modified by ibmi-agent-kit (2026): rewritten for the kit (install, profile folders, protected settings, mounting, the kit's execution guard hook, the run report
diff --git a/packages/core/src/credential.ts b/packages/core/src/credential.ts
--- a/packages/core/src/credential.ts
+++ b/packages/core/src/credential.ts
@@ -122,1 +122,0 @@
-const text = Buffer.concat(chunks).toString("utf8").replace(/^/, "").replace(/\r?\n$/, "");
diff --git a/.claude/commands/opsx/propose.md b/.claude/commands/opsx/propose.md
--- a/.claude/commands/opsx/propose.md
+++ b/.claude/commands/opsx/propose.md
@@ -162,1 +162,0 @@
-- Do NOT copy `<context>`, `<rules>`, `<project_context>` blocks into the artifact

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: ibmi-agent-kit: Fake instruction boundary tag (KillerSkills verdict caution 62)

## What the honeypot observed
KillerSkills ran **ibmi-agent-kit@r1** 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/glama/ibmi-agent-kit/r1

## Findings with a location
- `.claude/skills/openspec-propose/SKILL.md:166` · Fake instruction boundary tag — .claude/skills/openspec-propose/SKILL.md:166 · Fake instruction boundary tag
- `packages/screen/README.md:3` · HTML comment carrying imperative instructions to the agent — packages/screen/README.md:3 · HTML comment carrying imperative instructions to the agent
- `packages/core/src/credential.ts:122` · Hidden zero-width characters detected — packages/core/src/credential.ts:122 · Hidden zero-width characters detected — may conceal content
- `.claude/commands/opsx/propose.md:162` · Fake instruction boundary tag — .claude/commands/opsx/propose.md:162 · Fake instruction boundary tag · informational: not an instruction file

## Proposed change
The attached patch removes the 4 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._