Security audit
Security checks for vulnerabilities and agentic risk
This skill is a coherent internet speed-test and reporting tool, but users should understand that scheduled use creates recurring third-party network traffic and local logs.
Install only if you are comfortable running speed tests that contact external services, download up to the configured byte limits, upload about 5 MB to Cloudflare per run, and store timestamped results locally. Do not enable the cron or launchd schedule on metered, enterprise, or sensitive networks unless recurring traffic every 30 minutes is acceptable; review or delete ~/.isp-throttle-detective when you no longer need the history.
This second mismatch likewise indicates the skill's declared purpose does not match actual code behavior, including missing analysis/reporting features and different endpoint behavior. Such discrepancies can conceal what the skill really accesses or fails to do, undermining least privilege decisions and making review, consent, and monitoring less reliable.
This second mismatch likewise indicates the skill's declared purpose does not match actual code behavior, including missing analysis/reporting features and different endpoint behavior. Such discrepancies can conceal what the skill really accesses or fails to do, undermining least privilege decisions and making review, consent, and monitoring less reliable.
The skill advertises and orchestrates shell execution, network access, and persistent file logging, but it does not declare any explicit tool scope or permissions boundaries. This creates an overbroad execution surface where an agent may invoke sensitive capabilities without clear policy review, increasing the risk of unintended filesystem changes or network activity.
The manifest description says to use the skill when the user mentions phrases like "internet slow" and "internet speed," which are broad everyday expressions and could match many general troubleshooting conversations. The file does not provide exclusion conditions or tighter scope boundaries to clarify when the skill should not activate.
Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.
Copy the example config and customize:
mkdir -p ~/.isp-throttle-detective
cp assets/config.example.json ~/.isp-throttle-detective/config.json
The guide instructs users to schedule recurring speed tests every 30 minutes but does not warn that this will continuously generate network traffic to third-party endpoints and may consume bandwidth, affect data-capped plans, or create privacy-relevant telemetry. Because the purpose of the skill is to repeatedly contact external services, omitting this warning can lead users to unknowingly enable persistent outbound activity.
Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.
Create ~/Library/LaunchAgents/com.isp-throttle-detective.plist:
<?xml version="1.0" encoding="UTF-8"?>
Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.
Create ~/Library/LaunchAgents/com.isp-throttle-detective.plist:
<?xml version="1.0" encoding="UTF-8"?>
Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.
Create ~/Library/LaunchAgents/com.isp-throttle-detective.plist:
<?xml version="1.0" encoding="UTF-8"?>
Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.
Create ~/Library/LaunchAgents/com.isp-throttle-detective.plist:
<?xml version="1.0" encoding="UTF-8"?>
Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.
Create ~/Library/LaunchAgents/com.isp-throttle-detective.plist:
<?xml version="1.0" encoding="UTF-8"?>
Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.
<?xml version="1.0" encoding="UTF-8"?>
<!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd">
<plist version="1.0">
<dict>
<key>Label</key>
Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.
Load it:
launchctl load ~/Library/LaunchAgents/com.isp-throttle-detective.plist
Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.
crontab -e
# Add:
*/30 * * * * python3 /path/to/skills/isp-throttle-detective/scripts/speedtest.py | python3 /path/to/skills/isp-throttle-detective/scripts/log_result.py >> ~/.isp-throttle-detective/cron.log 2>&1
subprocess module calls execute external commands. Without careful input validation, this enables command injection.
def run_analysis(log_path: str, days: int) -> dict:
result = subprocess.run(
[sys.executable, ANALYZE_SCRIPT, "--log", log_path, "--days", str(days), "--json"],
capture_output=True, text=True
)
The manifest describes a speed-testing and reporting skill, which justifies network access, but it does not justify spawning external processes. Using subprocess.run to invoke curl adds an execution capability beyond the obvious needs of the stated purpose and broadens the runtime attack surface compared with using in-process HTTP libraries alone.
subprocess module calls execute external commands. Without careful input validation, this enables command injection.
"-A", "isp-throttle-detective/1.0",
url
]
result = subprocess.run(cmd, capture_output=True, text=True, timeout=timeout + 5)
if result.returncode not in (0, 63): # 63 = max-filesize exceeded (expected)
return {"speed_mbps": None, "error": result.stderr.strip() or f"curl exit {result.returncode}"}
parts = result.stdout.strip().split()
subprocess module calls execute external commands. Without careful input validation, this enables command injection.
def curl_available() -> bool:
try:
subprocess.run(["curl", "--version"], capture_output=True, timeout=5)
return True
except Exception:
return False
The function performs a real 5 MB upload to a third-party endpoint automatically, which transmits user network traffic externally without an explicit warning or opt-in at execution time. In a skill that users may run for diagnosis, this can create privacy, policy, or cost concerns, especially on metered, enterprise, or sensitive networks where outbound testing may be unexpected.
The skill persists test results under the user's home directory but does not clearly disclose that retention behavior in the user-facing description. Undisclosed local logging can surprise users, create privacy concerns, and leave historical network performance data on disk longer than expected.
The module usage docstring describes only direct log-file and day-range analysis options, but the code also accepts --config and reads an arbitrary JSON config file to override log_path at L234-L240. For a skill described as analyzing speed-test history, reading a separate configuration file is adjacent to its purpose but still an undocumented behavior mismatch between the file's stated interface and actual operations.
This code accepts a --config path and opens the specified file, which is a file-read operation affecting user data, but there is no prompt, comment, or usage text explaining that the script will read configuration content from disk. The only documented usage examples mention --log, --out, and --days, so the file access behavior is not disclosed in the visible interface.
When no output path is provided, the script silently writes a detailed report to a persistent directory under the user's home folder and then prints the report contents. In the context of a network-diagnostics skill, that report may contain timestamps, usage patterns, endpoint data, and filesystem paths, creating a privacy leak and unexpected local data retention if the user expected analysis-only behavior.
No suspicious patterns detected.