The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill references downloads from well-known and official sources including Docker images from the official Bun organization (oven/bun), GitHub Actions (oven-sh/setup-bun), and deployment tools for platforms like Fly.io and Vercel.
[COMMAND_EXECUTION]: Implementation patterns in SKILL.md and references/bun-essentials.md demonstrate and encourage the use of shell execution via Bun.shell ($), Bun.spawnSync, and Bun.spawn for tasks like file management, build processes, and deployment.
[PROMPT_INJECTION]: Eight reference files (including architecture-patterns.md, code-organization.md, and deployment-pipelines.md) contain hundreds of identical repetitive sections described as 'extremely dense technical specifications' for 'Staff+ Engineers'. This deceptive and voluminous content can lead to context drowning, potentially causing the agent to ignore critical instructions or safety guardrails.
[PROMPT_INJECTION]: The skill introduces an indirect prompt injection vulnerability surface (Category 8) through its provided code templates:
Ingestion points: Untrusted data enters the agent context via HTTP Request objects in src/index.ts and src/router.ts, as well as filename arguments in src/utils/file-storage.ts.
Boundary markers: Explicit delimiters or warnings to ignore instructions embedded in the ingested data are absent in the examples.
Capability inventory: The skill uses sensitive capabilities including Bun.spawnSync(['rm', path]) for file deletion and Bun.shell ($) for build and deployment scripts.
Sanitization: No validation or sanitization logic is present in the provided templates to prevent malicious input from influencing command execution or file operations.

bun