The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external sources that could theoretically contain malicious instructions designed to influence the agent's behavior.
Ingestion points: The skill fetches external data such as commit messages, branch names, and pull request information from GitHub Actions and Vercel deployments via CLI tools and API calls.
Boundary markers: There are no explicit instructions or delimiters defined to separate these untrusted external strings from the agent's system instructions.
Capability inventory: The skill has access to sensitive tools including Bash, WebFetch, TaskCreate, and Agent coordination, and can trigger deployment operations.
Sanitization: The instructions do not specify any sanitization, filtering, or escaping of the incoming metadata (e.g., commit messages) before they are displayed or processed by the agent.
[SAFE]: The skill implements strong operational guardrails and credential management practices.
Credentials for AWS and Vercel are resolved securely through environment variables or by querying the Doppler secret management tool, avoiding any hardcoded secrets.
High-impact actions, such as manual deployments, are explicitly gated by human-in-the-loop confirmation using AskUserQuestion.
Network operations are directed to well-known service APIs (e.g., Vercel API) for legitimate operational data retrieval and follow-up actions.

ops-deploy