The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill uses bunx algal-skills to fetch and execute a package directly from the NPM registry at runtime. While NPM is a standard registry, the package itself originates from a source not explicitly categorized as a trusted vendor.
[DYNAMIC_EXECUTION]: The skill operates by spawning processes based on 'router manifests' proposed by an agent. This involves dynamic execution of logic or configurations generated at runtime.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a surface for indirect injection as it ingests 'candidate router manifests' proposed by an agent. If these manifests contain malicious instructions or parameters, they could influence the behavior of the 'children' processes or the 'router-live' consumer.
Ingestion points: Agent-proposed manifests in the router-habitat environment.
Boundary markers: None explicitly defined in the instructions to separate manifest data from control logic.
Capability inventory: The skill utilizes the exec tool to run the evolution driver and spawn candidate manifests.
Sanitization: The skill relies on a fitness scorer (expr) to promote manifests, which measures performance but does not necessarily validate instructions for security.
[COMMAND_EXECUTION]: The skill is designed to execute shell commands via bunx to run the evolution driver, which triggers remote code execution patterns.

algal-evolve