The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection via external data sources. Ingestion points include GitLab issue descriptions, comments, and GitHub issue content. Boundary markers are absent in the sub-agent prompt templates, meaning external data is interpolated directly into {context} and {all_opening_statements}. The skill has capabilities to write back to issues and spawn sub-agents. No sanitization or filtering is performed on the ingested content, which could allow an attacker to influence the persona debate or meeting summary by embedding instructions in an issue comment.- [EXTERNAL_DOWNLOADS]: Fetches issue details and comments from GitLab and GitHub using the gitlab-mcp tools and the gh CLI. These operations are core to the skill's functional purpose of analyzing project-specific context from well-known version control services.

meeting-personas