The Agent Skills Directory

[SAFE]: The skill's instructions and implementation align with its stated purpose of media licensing verification. The logic is deterministic and focused on identifying specific credit patterns associated with NASA, ESA, and other observatories.
[INDIRECT_PROMPT_INJECTION]: The skill has a data ingestion surface as it evaluates external content from user-provided URLs and credit strings.
Ingestion points: The asset input (processed via the WebFetch tool) and the credit text input are the primary vectors where untrusted data enters the agent context.
Boundary markers: The instructions do not include explicit delimiters (e.g., XML tags or triple quotes) or 'ignore embedded instructions' directives to wrap the fetched content.
Capability inventory: The skill is limited to the WebFetch and Read tools. It lacks high-risk capabilities such as file system writes, network exfiltration (POST/PUT), or shell command execution.
Sanitization: The skill does not explicitly sanitize the external content, but the risk is mitigated by the restricted output format, which is a static JSON contract for licensing metadata with no path to executable sinks.

rights-check-nasa-esa