Propose a patch

Evidence

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

diff --git a/safe_path.py b/safe_path.py
--- a/safe_path.py
+++ b/safe_path.py
@@ -12,1 +12,0 @@
-`.faf` / `.fafm` context file. Blocks /etc/passwd, ~/.ssh/id_rsa,
@@ -13,1 +12,0 @@
-~/.aws/credentials, .env, etc. regardless of directory.
diff --git a/README.md b/README.md
--- a/README.md
+++ b/README.md
@@ -46,1 +46,0 @@
-**faf_gemini and faf_agents keep your hand-written GEMINI.md and AGENTS.md, and tell you what they couldn't write.**
diff --git a/inject.py b/inject.py
--- a/inject.py
+++ b/inject.py
@@ -136,1 +136,0 @@
-if existing.lstrip("").lstrip().startswith(FAF_METASTAMP) or _has_unpaired_marker(existing, start, end):
@@ -250,1 +249,0 @@
-bom = "" if existing.startswith("") else ""
diff --git a/PKG-INFO b/PKG-INFO
--- a/PKG-INFO
+++ b/PKG-INFO
@@ -88,1 +88,0 @@
-**faf_gemini and faf_agents keep your hand-written GEMINI.md and AGENTS.md, and tell you what they couldn't write.**
diff --git a/gemini_faf_mcp.egg-info/PKG-INFO b/gemini_faf_mcp.egg-info/PKG-INFO
--- a/gemini_faf_mcp.egg-info/PKG-INFO
+++ b/gemini_faf_mcp.egg-info/PKG-INFO
@@ -88,1 +88,0 @@
-**faf_gemini and faf_agents keep your hand-written GEMINI.md and AGENTS.md, and tell you what they couldn't write.**
diff --git a/server.py b/server.py
--- a/server.py
+++ b/server.py
@@ -76,1 +76,0 @@
-"project.faf is the source: faf_gemini and faf_agents write GEMINI.md and AGENTS.md from it, leaving hand-written content alone.",
@@ -350,1 +349,0 @@
-"""Export and write AGENTS.md from a .faf file (non-destructive).

Zero-context hunks: apply with git apply --unidiff-zero fix.patch on the tagged source of 3.1.2. 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: gemini-faf-mcp: References a credential file in the home directory (KillerSkills verdict caution 65)

## What the honeypot observed
KillerSkills ran **[email protected]** in an isolated sandbox (decoy credentials, recording proxy, syscall trace; engine 1.6) and scored it **CAUTION 65/100**. Summary: no network, touches nothing outside the project, writes no memory.

- See the findings below.

Full trace and findings: https://killerskills.ai/skills/pypi/gemini-faf-mcp/3.1.2

## Findings with a location
- `safe_path.py:12` · References a credential file in the home directory — safe_path.py:12 · References a credential file in the home directory
- `README.md:46` · Instructs the agent to modify its persistent memory or instruction files — README.md:46 · Instructs the agent to modify its persistent memory or instruction files · informational: not an instruction file
- `safe_path.py:13` · References a credential file in the home directory — safe_path.py:13 · References a credential file in the home directory
- `inject.py:250` · Hidden zero-width characters detected — inject.py:250 · Hidden zero-width characters detected — may conceal content
- `inject.py:136` · Hidden zero-width characters detected — inject.py:136 · Hidden zero-width characters detected — may conceal content
- `PKG-INFO:88` · Instructs the agent to modify its persistent memory or instruction files — PKG-INFO:88 · Instructs the agent to modify its persistent memory or instruction files · informational: not an instruction file
- `gemini_faf_mcp.egg-info/PKG-INFO:88` · Instructs the agent to modify its persistent memory or instruction files — gemini_faf_mcp.egg-info/PKG-INFO:88 · Instructs the agent to modify its persistent memory or instruction files · informational: not an instruction file
- `server.py:76` · Instructs the agent to modify its persistent memory or instruction files — server.py:76 · Instructs the agent to modify its persistent memory or instruction files · informational: not an instruction file
- `server.py:350` · Instructs the agent to modify its persistent memory or instruction files — server.py:350 · Instructs the agent to modify its persistent memory or instruction files · informational: not an instruction file

## Proposed change
The attached patch removes the 9 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._