The Agent Skills Directory

Remote Script Execution: The skill includes instructions to download and execute a shell script from the official astral.sh domain to install the uv Python package manager. While this is a standard installation method for this well-known developer tool, executing remote scripts directly in the shell using the curl | sh pattern is a security consideration.
Indirect Prompt Injection Surface: The skill is designed to ingest and analyze untrusted data by reading agent logs and execution traces via commands like agentcore logs and agentcore traces get.
Ingestion points: External data enters the agent's context through log search and trace retrieval operations defined in SKILL.md.
Boundary markers: The instructions do not specify explicit delimiters or "ignore embedded instructions" warnings for the ingested log content.
Capability inventory: The skill possesses the Bash tool and can execute various system commands, including aws and agentcore CLI operations.
Sanitization: No specific filtering or sanitization of the retrieved log content is performed before the agent processes it.
Privilege Escalation: For environment troubleshooting, the skill suggests using sudo to manage the Docker service. These administrative commands are necessary for service management but represent an escalation of privileges.
External Software Management: The skill facilitates the installation and uninstallation of various packages via npm, pip, and uv. While these involve trusted vendor resources and well-known developer tools, managing external software introduces dependencies on third-party registries.

agents-debug