The Agent Skills Directory

[COMMAND_EXECUTION]: The skill includes Python and Shell examples that execute external command-line tools such as sonar-scanner, semgrep, codeql, and trufflehog via subprocess.run and standard shell execution. These operations are aligned with the skill's core purpose of automating code quality and security reviews.
[EXTERNAL_DOWNLOADS]: The CI/CD workflow examples reference trusted GitHub Actions, specifically actions/checkout@v4 and actions/github-script@v7. It also references official tools and libraries from well-known organizations including Anthropic and OpenAI, which are treated as safe sources.
[PROMPT_INJECTION]: The skill contains an indirect prompt injection vulnerability (Category 8).
Ingestion points: Untrusted data from pull request diffs and descriptions are read directly into the agent's context within SKILL.md.
Boundary markers: The prompt templates do not utilize clear delimiters or instructions to ignore embedded commands within the ingested code or descriptions.
Capability inventory: The orchestrator has the capability to run shell commands, interact with the GitHub API to post comments, and perform network requests to AI model providers.
Sanitization: There is no evidence of sanitization or escaping of the code_diff or pr_description variables before they are interpolated into the final prompt strings.

code-review-ai-ai-review