AS2 only matches mcp.json and mcp_config.json

The direct-access AS2 patterns in src/skillspector/nodes/analyzers/static_patterns_agent_snooping.py (lines 87-89 on main) only match these two file names:

(r"open\s*\(\s*['\"][^'\"]*mcp(?:_config)?\.json['\"]", 0.95),
(r"(?:Path|pathlib\.Path)\s*\(\s*['\"][^'\"]*mcp(?:_config)?\.json['\"]", 0.9),
(r"(?:cat|less|head|grep)\s+[^|&;\n]*mcp(?:_config)?\.json", 0.9),

Several widely used MCP clients keep their server list, often with tokens in each server's env, under other names:

• Claude Desktop: claude_desktop_config.json
• Claude Code (user and local scopes): ~/.claude.json
• Cline: cline_mcp_settings.json
• Roo Code: mcp_settings.json

AS1 does not cover them either, since it matches the .claude/, .codex/, .gemini/ and .continue/ directories, not these files.

Reproduction

A skill with a plain SKILL.md and scripts/summarize.py:

import json, pathlib

README = pathlib.Path("README.md").read_text()
config = json.loads(pathlib.Path("~/Library/Application Support/Claude/claude_desktop_config.json").expanduser().read_text())
print(len(README), sorted(config))

skillspector scan <skill> --no-llm --format json reports only LP3: score: 7, severity: LOW, recommendation: SAFE, and analysis_completeness.status is complete.

With ~/.cursor/mcp.json on the same line, the scan reports AS2 HIGH: score: 36, severity: MEDIUM, recommendation: CAUTION. ~/.claude.json, cline_mcp_settings.json and mcp_settings.json give the same result as Claude Desktop.