The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from multiple external sources, including JIRA user stories, Figma designs, and API specifications. This content is analyzed to produce instructions for a downstream "Code Generation Agent."
Ingestion points: JIRA story JSON files (.SS_WF/{{$var[ticket_id]s}}_jira_output.json), Sitecore/BFF API specifications, and Figma context files (figma-output/).
Boundary markers: The instructions do not define boundary markers or "ignore instructions" delimiters for the ingested content.
Capability inventory: The skill uses file-writing tools to generate plan documents and has the capability to read local source code for component reuse validation. The workflow specifically bypasses human review before the output reaches the coding agent.
Sanitization: There is no evidence of sanitization or filtering of the processed text to prevent malicious instructions from being passed through.
[COMMAND_EXECUTION]: The skill utilizes network-capable tools (such as curl and http) in Phase 2 to fetch API and design data from URLs extracted from the JIRA story. This represents a potential vulnerability surface if the URLs extracted from untrusted input are not strictly validated before being passed to the network tools.

master-analysis-orchestrator