Propose a patch

Evidence

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

diff --git a/src/bud_relay/cli.py b/src/bud_relay/cli.py
--- a/src/bud_relay/cli.py
+++ b/src/bud_relay/cli.py
@@ -133,1 +133,0 @@
-"# bud-relay secrets. Keep this file private; it is gitignored.",
diff --git a/src/bud_relay/agents/claude_cli.py b/src/bud_relay/agents/claude_cli.py
--- a/src/bud_relay/agents/claude_cli.py
+++ b/src/bud_relay/agents/claude_cli.py
@@ -97,1 +97,0 @@
-# The relay's variables a run may see, when set: none of them is a relay secret.
@@ -113,1 +112,0 @@
-# Set in every run: no auto-update or telemetry, no CLAUDE.md from anywhere, and an API error
@@ -571,1 +569,0 @@
-"repo and a shell) is planned for 0.3.0, in a separate container with no relay secrets"
diff --git a/CONTRACT.md b/CONTRACT.md
--- a/CONTRACT.md
+++ b/CONTRACT.md
@@ -195,1 +195,0 @@
-gains a line `Your user's own rules: <instructions>` (`tools.answering_rules`). The relay

Zero-context hunks: apply with git apply --unidiff-zero fix.patch on the tagged source of 0.2.1. 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: bud-relay: Credential pass-through configuration (KillerSkills verdict caution 63)

## What the honeypot observed
KillerSkills ran **[email protected]** in an isolated sandbox (decoy credentials, recording proxy, syscall trace; engine 1.7) 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/bud-relay/0.2.1

## Findings with a location
- `src/bud_relay/cli.py:133` · Credential pass-through configuration — src/bud_relay/cli.py:133 · Credential pass-through configuration — agent relays sensitive credentials
- `src/bud_relay/agents/claude_cli.py:97` · Credential pass-through configuration — src/bud_relay/agents/claude_cli.py:97 · Credential pass-through configuration — agent relays sensitive credentials
- `src/bud_relay/agents/claude_cli.py:571` · Credential pass-through configuration — src/bud_relay/agents/claude_cli.py:571 · Credential pass-through configuration — agent relays sensitive credentials
- `src/bud_relay/agents/claude_cli.py:113` · Instructs the agent to modify its persistent memory or instruction files — src/bud_relay/agents/claude_cli.py:113 · Instructs the agent to modify its persistent memory or instruction files · informational: not an instruction file
- `CONTRACT.md:195` · Fake instruction boundary tag — CONTRACT.md:195 · Fake instruction boundary tag · informational: not an instruction file

## 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._