The Agent Skills Directory

[SAFE]: The skill implements a structured PR review process using co-located instruction files (_review-core.md, agents/devils-advocate.md, agents/jira-completeness.md).
[SAFE]: It utilizes standard system tools like git and gh (GitHub CLI) to fetch diffs and metadata. Environment detection logic appropriately distinguishes between restricted CI environments and local environments with network access.
[SAFE]: The JIRA integration uses the official mcp__atlassian tools to fetch issue details from a specified Cloud ID, which is a standard pattern for MCP-based Jira interactions.
[SAFE]: The skill implements a 'Devil's Advocate' mode and 'Full' mode that spawns sub-agents for specialized analysis. These sub-agents are constrained by the same safety and methodology guidelines as the primary agent.
[SAFE]: No evidence of prompt injection, data exfiltration, or unauthorized command execution was found. The instructions emphasize flagging only actionable issues introduced by the PR and explicitly exclude non-relevant concerns like style issues or unmodified code.

pr-review