The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill uses npx @useclawlink/cli to download and run the vendor's command-line tool directly from the npm registry. This is the intended delivery method for the integration.
[COMMAND_EXECUTION]: The skill relies on shell commands (npx) to perform login, connection, and data operations within the New Relic environment. It stores authentication tokens locally at ~/.clawlink/credentials.json.
[INDIRECT_PROMPT_INJECTION]: The skill processes user-defined strings and JSON inputs to interact with the New Relic API, creating a vulnerability surface where untrusted data could influence agent actions.
Ingestion points: Action descriptions provided to npx @useclawlink/cli actions and JSON input provided to the --input flag in SKILL.md.
Boundary markers: None identified; the skill passes raw strings directly to the CLI tool.
Capability inventory: Includes the ability to read and write New Relic data, modify alert policies, and manage cloud integrations via CLI subprocesses.
Sanitization: No evidence of input validation or escaping for the user-provided action descriptions or JSON payloads is present in the skill instructions.

new-relic