The Agent Skills Directory

[Prompt Injection] (MEDIUM): The skill creates a vulnerability to indirect prompt injection. It instructs the agent to run speakeasy agent context and use its output as the primary source of truth for subsequent actions.
Ingestion points: Output of speakeasy agent context in SKILL.md.
Boundary markers: Absent; no instructions are provided to the agent to treat the CLI output as untrusted data.
Capability inventory: Ability to execute various speakeasy CLI commands and send data via feedback (SKILL.md).
Sanitization: None; the skill lacks validation or escaping for the data received from the CLI.
[Data Exfiltration] (LOW): The speakeasy agent feedback command is used to transmit data to an external service. While documented as a feedback mechanism, it represents a routine outbound data flow.
[Command Execution] (LOW): The skill is built around executing local CLI commands. It does not attempt to download or install new software, but relies on the existing speakeasy binary.

speakeasy-context