The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes bash commands for environmental verification and project lifecycle management.
Evidence includes project initialization checks (ls -la package.json), dependency management (bun add), and process control for port cleanup (lsof -ti:3000 combined with kill -9).
[DYNAMIC_EXECUTION]: Implements dynamic module loading to facilitate live updates without full application restarts.
The Hot Module Replacement (HMR) logic in server.ts and hmr-examples.md uses dynamic import() statements with computed paths (cache-busting timestamps like ?t=${Date.now()}) to reload updated components and routes at runtime.
[INDIRECT_PROMPT_INJECTION]: The skill defines a server architecture that ingests external data, which represents an attack surface for indirect prompt injection if processing instructions are embedded in that data.
Ingestion points: Server handlers process untrusted data from request.url, c.req.json() payloads, and local file system paths via Bun.file(filePath) in SKILL.md.
Boundary markers: None identified in the provided templates to distinguish between data and control instructions.
Capability inventory: The skill uses Bash, Write, and Read tools to create and manage the server environment.
Sanitization: The code templates do not include explicit sanitization or validation of the input paths or payloads before they are processed by the server logic.
[PRIVILEGE_ESCALATION]: The skill suggests using local development tools that interact with system-level configurations.
Documentation in SKILL.md guides the user to use mkcert for generating locally-trusted certificates, which involves modifying the system's root trust store to enable local HTTPS development.

bun-dev-server