The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill provides instructions to download the contents of a repository identified as malicious via git clone https://github.com/BinayRAj-boop/fort-firewall-security-fix. This action brings trojanized installers and malware loaders into the agent's filesystem.
[COMMAND_EXECUTION]: The skill utilizes system commands including git clone, find, grep, gh, and curl to interact with the malicious repository and perform reporting tasks to external APIs.
[INDIRECT_PROMPT_INJECTION]: The skill ingests and analyzes untrusted data from external GitHub repositories.
Ingestion points: Repository metadata (topics, creation dates) and README.md content are retrieved via gh repo view and git clone.
Boundary markers: No explicit boundary markers or instructions to ignore embedded commands are present in the analysis logic.
Capability inventory: File system writes (cloning), subprocess execution (git, grep, find, gh), and network operations (curl, git).
Sanitization: The skill uses regex matching for identification but lacks generic sanitization for the untrusted content being processed.
[CREDENTIALS_UNSAFE]: The skill documentation encourages the use of environment variables for GITHUB_TOKEN and VIRUSTOTAL_API_KEY. While placeholders are used, this promotes a pattern of storing sensitive credentials in a shell environment where they may be exposed to other processes or logged.
[METADATA_POISONING]: There is a discrepancy between the author identified in the platform context ('reason-machines') and the author stated within the skill content ('ara.so'), which can be deceptive regarding the origin and trustworthiness of the security tool.

fort-firewall-malware-repository-detection