github-issues
Pass
Audited by Gen Agent Trust Hub on Oct 23, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to read and process external, untrusted content from GitHub issues, comments, and project items. This content is then used to perform automated actions, creating a risk that malicious instructions embedded in issue data could influence the agent's behavior.
- Ingestion points: Multiple tools such as
mcp__github__issue_readandmcp__github__list_issues(SKILL.md), and GraphQL queries for search and projects (references/search.md, references/projects.md) ingest external strings. - Boundary markers: The skill does not define strict boundary markers or sanitization logic when interpolating external issue content into further commands or prompts.
- Capability inventory: The skill has high-privilege write access through MCP tools like
mcp__github__issue_write,mcp__github__projects_write, and thegh apiCLI interface, allowing for modifications to repositories and project boards. - Sanitization: No explicit sanitization or escaping of external content is mandated before it is used in write operations.
- [DYNAMIC_EXECUTION]: The documentation provides a script template that utilizes
puppeteer-corefor taking screenshots, which involves local execution and filesystem interaction. - Evidence:
references/images.mdcontains a JavaScript snippet that usesrequire('puppeteer-core')andpuppeteer.launchwith a hardcoded local executable path (/Applications/Google Chrome.app/Contents/MacOS/Google Chrome) and writes results to the local filesystem.
Audit Metadata