The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from pull request diffs and metadata, which could contain malicious instructions from external contributors.
Ingestion points: Pull request titles, bodies, and diff content are ingested via gh pr view and gh pr diff in SKILL.md (Step 1).
Boundary markers: The instructions lack explicit delimiters or markers to isolate untrusted pull request content when it is passed to sub-agents or processed by the primary agent.
Capability inventory: The skill has access to Bash(gh *), Bash(git *), Agent, Skill, and CallMcpTool, allowing for file system interaction, orchestration of other skills, and internal network requests via the MCP.
Sanitization: There are no specified procedures for sanitizing or escaping the data retrieved from pull requests before it is processed.
[EXTERNAL_DOWNLOADS]: The skill retrieves documentation guidelines and pull request data from official and trusted sources.
Evidence: SKILL.md fetches information from elastic.co, elastic.dev, and github.com.
[COMMAND_EXECUTION]: The skill uses command-line tools to interact with the repository and verify the state of the working tree.
Evidence: The skill executes gh pr view, gh pr diff, git rev-parse, and git fetch to resolve review targets and compare against base branches.

docs-review-pr