The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes potentially untrusted data from host logs and command outputs. It addresses this surface with a mandatory evidence chain: 1) Ingestion points include output from observe-agent status, observe datastream view, and system logs; 2) Boundary markers are provided by a wrap helper that uses cryptographically random nonces; 3) Capability inventory includes shell command execution and Observe CLI operations; 4) Sanitization is performed via a sed and tr pipeline to remove ANSI escape sequences and non-printable characters.
[COMMAND_EXECUTION]: The skill instructs the agent to perform several shell-based operations to interact with the Observe platform. Evidence includes commands such as observe auth status, observe auth login, and observe datastream view.
[PRIVILEGE_ESCALATION]: Administrative privileges are requested for host setup tasks. Evidence includes the use of sudo hostnamectl and the installation of the observe-agent as a systemd service, which are necessary for the skill's stated purpose.
[DYNAMIC_EXECUTION]: The skill dynamically updates its own instructions and loads referenced modules. Evidence includes the mandatory requirement to run observe skill view --content and --path to load the latest skill logic and supporting documents at runtime.

deploy-linux-host-explorer