The Agent Skills Directory

[PROMPT_INJECTION]: The skill presents an indirect prompt injection surface as it ingests and processes untrusted data from the repository being analyzed.
Ingestion points: Local codebase files are read during the Step 1 (Explore) process in SKILL.md.
Boundary markers: No explicit boundary markers or instructions to ignore embedded directives are used when reading files.
Capability inventory: The skill has the capability to create GitHub issues using gh issue create and execute arbitrary architectural analysis via the Agent tool.
Sanitization: No sanitization or filtering of the ingested code content is specified before it is processed by sub-agents or included in GitHub issues.

refactor