The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from tickets, feedback, and attachments which could contain malicious instructions designed to influence the agent.
Ingestion points: Data is collected from the CawPlan system using tickets get, execution summary, and testpoints list commands.
Boundary markers: The skill explicitly references references/UNTRUSTED_CONTENT.md and includes instructions to treat ticket data as non-executable information.
Capability inventory: The skill has the ability to read ticket data and write to the QA reports portal via CLI tools.
Sanitization: A local script scripts/validate_report_details.js validates HTML structure and link formatting, though it does not provide comprehensive sanitization of the ticket-derived text itself.
[DYNAMIC_CONTEXT_INJECTION]: The skill uses a dynamic execution placeholder in the main instruction file to perform a configuration check.
Evidence: The command !cawplan skill check`` is present in SKILL.md`.
Analysis: This is a diagnostic command that runs at skill load time. It appears to be a benign environment check and does not handle sensitive data or user-supplied arguments.
[COMMAND_EXECUTION]: The skill executes a local Node.js script and several vendor CLI tools as part of its workflow.
Evidence: Execution of node scripts/validate_report_details.js and various cawplan commands like tickets get and qa-reports create.
Analysis: The script is a local utility used to ensure that generated HTML is compact and formatted correctly for the destination portal. The CLI tools are scoped to the task of report generation and data collection.

cawplan-ticket-report-generate