The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests data from several sources that could potentially be influenced by external actors, such as trading position data, decision logs, and alerts from the OpenClaw Gateway.
Ingestion points: Data is fetched via prism status in SKILL.md and scripts/prism-checkin.sh, local HTTP requests to http://127.0.0.1:18790 for positions and decisions, and received alerts through the gateway defined by AGENT_GATEWAY_URL.
Boundary markers: The instructions do not specify any delimiters or warnings to ignore instructions embedded in the ingested data before it is processed by the AI agent.
Capability inventory: The skill has the ability to execute the prism CLI on the host and perform local network operations via curl to manage the trading agent.
Sanitization: There is no evidence of sanitization, validation, or escaping performed on the ingested data before it is interpolated into the agent context.
[COMMAND_EXECUTION]: The skill relies on the execution of the prism CLI tool on the host system to perform status checks, setup operations, and agent management. This is the intended behavior for operating the autonomous agent, but it involves the execution of external binaries.

prism-openclaw