flywheel-admin
Fail
Audited by Gen Agent Trust Hub on Jul 26, 2026
Risk Level: CRITICALREMOTE_CODE_EXECUTIONEXTERNAL_DOWNLOADSCREDENTIALS_UNSAFEDATA_EXFILTRATIONCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [REMOTE_CODE_EXECUTION]: The skill instructs the agent to download and execute scripts directly from the internet by piping
curloutput tobash, a high-risk pattern for arbitrary code execution. - Found in
SKILL.md:curl -fsSL https://raw.githubusercontent.com/Dicklesworthstone/agentic_coding_flywheel_setup/main/install.sh | bash -s -- --easy-mode --skip-ubuntu-upgrade. - Found in
references/known-quirks.md:curl -fsSL https://raw.githubusercontent.com/Dicklesworthstone/mcp_agent_mail_rust/main/install.sh | bash -s -- --yes --easy-mode --migrate --verify. - [EXTERNAL_DOWNLOADS]: Fetches multiple resources from an untrusted external GitHub repository (
Dicklesworthstone), including installation scripts and stack components. - [CREDENTIALS_UNSAFE]: Accesses and manages sensitive local configuration files and security keys required for remote access.
- References
~/.ssh/configfor managing host connection details. - Explicitly references the private key path
~/.ssh/id_ed25519_flywheelfor SSH identity inreferences/connect.md. - Sources local environment configuration from
~/.admin/.env. - [DATA_EXFILTRATION]: Provides procedures to transfer data between the local machine and remote VPS instances using
scp. - Mirrors the local skill bundle to remote servers via
scp -r "$SKILLS_SRC/." flywheel-N-oci:~/skills-staging/. - Downloads project artifacts from remote hosts via
scp -r flywheel-N-oci:/data/projects/<name>/dist ./. - [COMMAND_EXECUTION]: Uses complex shell command sequences involving
ssh,jq, andntmto manage fleet operations and inspect remote systems. - [PROMPT_INJECTION]: The skill possesses a vulnerability surface for indirect prompt injection by processing external markdown files.
- Ingestion points: Reads
PRD.mdandAGENTS.mdand transfers them to remote flywheels. - Boundary markers: Missing markers to distinguish between system instructions and embedded file content.
- Capability inventory: Uses
ssh,ntm spawn, andntm sendto direct remote agents based on the content of these files. - Sanitization: No filtering or sanitization of external file content before processing.
Recommendations
- HIGH: Downloads and executes remote code from: https://raw.githubusercontent.com/Dicklesworthstone/agentic_coding_flywheel_setup/main/install.sh - DO NOT USE without thorough review
- AI detected serious security threats
Audit Metadata