The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the agent to execute commands using npx @leoflores/datadog-cli across multiple files (SKILL.md, references/dashboards.md, etc.). Running commands via npx without explicit version pinning dynamically fetches and executes code from the external npm registry at runtime, presenting a supply chain risk. Furthermore, there is a discrepancy between the repository documented in README.md (github.com/leonardocouy/datadog-cli) and the npm scope/package used in the actual execution (@leoflores/datadog-cli), indicating an unverified package source.
[INDIRECT_PROMPT_INJECTION]: The skill introduces an indirect prompt injection vulnerability surface by reading untrusted application logs and telemetry data.
Ingestion points: Data enters the agent's context through log search and real-time tailing commands (npx @leoflores/datadog-cli logs search and logs tail) defined in SKILL.md and references/logs-commands.md.
Boundary markers: There are no boundary markers or explicit isolation instructions defined to prevent the agent from executing instructions embedded within log messages.
Capability inventory: The agent possesses full shell execution capabilities via the npx tool and modification capabilities on external dashboards (dashboards update, dashboards delete).
Sanitization: No sanitization, filtering, or validation is performed on the incoming log data before it is processed.

datadog-cli