The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill clones an external repository (https://github.com/KeygraphHQ/shannon.git) and proceeds to execute the downloaded ./shannon script. This allows for the execution of arbitrary code from a source not recognized as a trusted vendor.
[COMMAND_EXECUTION]: User-provided strings for the target URL and repository name are directly interpolated into shell commands (e.g., ./shannon start URL={TARGET_URL} REPO={REPO_NAME}). A malicious user or attacker could provide input containing shell metacharacters (like ;, &, or |) to execute unintended commands on the host system.
[CREDENTIALS_UNSAFE]: The skill checks for the presence of and instructs users on how to set sensitive credentials, including ANTHROPIC_API_KEY and AWS access keys, which are then used by the tool.
[EXTERNAL_DOWNLOADS]: The skill performs an automated download of the Shannon toolset from GitHub during the setup phase.
[INDIRECT_PROMPT_INJECTION]: The agent is instructed to read and interpret pentest reports produced by external processes. If these reports contain adversarial instructions, they could potentially influence the agent's subsequent actions, such as recommending malicious code remediations.

shannon