The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill uses npx @reticlehq/server@latest to fetch and execute its server component from the npm registry. As the package is maintained by the skill's author (reticlehq) and uses a well-known registry, this is standard functionality.
[COMMAND_EXECUTION]: The skill instructs the agent to run various setup and diagnostic commands such as npx @reticlehq/server init and npx @reticlehq/server doctor. These commands are used for project initialization and troubleshooting.
[INDIRECT_PROMPT_INJECTION]: The skill has an attack surface for indirect prompt injection because it tools read live application data (DOM, network requests, console logs).
Ingestion points: Untrusted data enters the context via reticle_look, reticle_observe, and other monitoring tools (SKILL.md).
Boundary markers: The instructions mandate that the agent paste the reticleResult block verbatim into its response to prevent hallucination or misinterpretation.
Capability inventory: The skill can execute shell commands via npx and perform file system writes during the init phase (SKILL.md).
Sanitization: The skill explicitly warns the agent: 'Never put app source, secrets or user data in a report.'
[PRIVILEGE_ESCALATION]: The init command modifies agent configurations to register the Reticle server. However, the skill explicitly instructs the agent to only run this with human agreement and never to bypass or suppress host approval prompts.

install-and-verify