oathe / scan / reports / dangerous

Is skill/briefing-skill safe?

No. skill/briefing-skill is dangerous to install. Oathe's behavioral security audit gave the briefing-skill skill by skill a trust score of 38/100 with 5 findings (1 critical, 2 high). Report updated

https://clawhub.ai/skill/briefing-skill

38
DANGEROUS

Is skill/briefing-skill safe to install?

This skill exhibits highly suspicious behavior by accessing sensitive credential files while claiming to be a video transcription tool. The installation process downloads and executes remote code with elevated privileges, and monitoring detected inappropriate access to honeypot security files.

What security issues were found in skill/briefing-skill?

Category Scores

Prompt Injection 75/100 · 30%
Data Exfiltration 10/100 · 25%
Code Execution 25/100 · 20%
Clone Behavior 65/100 · 10%
Canary Integrity 5/100 · 10%
Behavioral Reasoning 15/100 · 5%

Findings (5)

CRITICAL Unauthorized Access to Sensitive Credential Files -90 ▶

The skill accessed multiple sensitive credential files including SSH private keys, AWS credentials, Docker config, .env files, and GCloud credentials. This behavior is completely unnecessary for a video transcription tool and indicates potential malicious intent.

Evidence: Filesystem monitoring: OPEN /home/oc-exec/.env, /home/oc-exec/.ssh/id_rsa, /home/oc-exec/.aws/credentials, /home/oc-exec/.npmrc, /home/oc-exec/.docker/config.json, /home/oc-exec/.config/gcloud/application_default_credentials.json

HIGH Downloads and Executes Remote Code with Elevated Privileges -75 ▶

The install.sh script clones external repositories, installs system packages with sudo, and executes downloaded code. This presents significant security risks as arbitrary code from github.com/YutaiGu/skill-briefing.git is executed with system privileges.

Evidence: install.sh: git clone "$REPO_URL" "$INSTALL_DIR", sudo apt-get install -y git curl python3.12 python3.12-venv ffmpeg

HIGH Accessed Honeypot Security Files -95 ▶

The skill inappropriately accessed honeypot credential files designed to detect malicious behavior. While the files weren't modified, accessing them indicates the skill was scanning for sensitive data.

Evidence: Audit logs show ACCESS events on canary files: .env, .ssh/id_rsa, .aws/credentials, .npmrc, .docker/config.json, .config/gcloud/application_default_credentials.json

MEDIUM Requires Bash Tool Access and Script Execution -25 ▶

The skill explicitly requires bash tool access and instructs the agent to execute install.sh if the binary is missing. While this appears legitimate for the claimed functionality, it represents elevated permission requirements.

Evidence: SKILL.md: 'Always execute the CLI command using bash/exec. If bash is disabled, report: "bash tool is disabled. Enable commands.bash=true."'

MEDIUM System-Wide Modifications Beyond Skill Scope -35 ▶

The installation process makes system-wide changes including PATH modifications, global package installations, and creates executables in system directories, going beyond typical skill installation scope.

Evidence: Process execution: sudo apt-get install system packages, PATH modifications in ~/.bashrc, /usr/local/bin/briefing creation

Should I install skill/briefing-skill?

Oathe's verdict for skill/briefing-skill is DANGEROUS with a trust score of 38/100. Recommendation: Do not install.