The Agent Skills Directory

[SAFE]: The skill is primarily instructional, focusing on best practices for secret management and regional endpoint configuration. It explicitly warns against anti-patterns such as committing API keys to source control or exposing them to client-side code.
[EXTERNAL_DOWNLOADS]: The skill references official documentation and API guides from AssemblyAI's public domain for technical verification.
[INDIRECT_PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection as it processes files within a repository. However, risk is mitigated by explicit boundaries and restricted tool capabilities.
Ingestion points: Repository code and configuration files accessed via Read, Glob, and Grep (SKILL.md).
Boundary markers: Explicit "Approval Boundaries" and "Tool Discipline" sections define limits on agent actions and credential handling.
Capability inventory: The skill's scope is restricted to local file operations (Read, Glob, Grep, Write, Edit) and lacks network or shell execution capabilities in the provided tool scope.
Sanitization: The instructions mandate the exclusion of credentials and customer-derived content from output, and require validation that sensitive fields are absent from evidence.

assemblyai-install-auth