The Agent Skills Directory

[Identity and Permission Verification]: The skill implements a robust pre-flight check that verifies the active GitHub account (gh auth status) and ensures the token has the necessary permissions (push or triage) before attempting any operations. This prevents the agent from performing actions as an unintended user or failing cryptically due to insufficient scopes.- [Dry-Run and Approval Workflow]: A significant security feature is the mandatory dry-run plan. The skill is instructed to never perform mutations (requesting reviewers) until the user has reviewed a detailed plan and provided explicit approval. This human-in-the-loop requirement mitigates risks associated with automated batch operations.- [Secure Argument and Query Handling]: The instructions emphasize building search qualifiers from validated repository and team values rather than string literals or direct user input. This practice, combined with proper form-encoding via the gh CLI, effectively prevents command injection and ensures the agent only operates on the intended resources.- [Configuration Integrity]: The skill uses a JSON schema (reviewers.schema.json) and jq validation to ensure the integrity of the reviewers.json configuration file. By setting additionalProperties: false in the schema, it prevents the introduction of unauthorized metadata or free-text fields that could be used for prompt injection or storing sensitive data.- [Least Privilege Data Handling]: The skill is designed to only request the minimum information needed (read:org for team rosters) and focuses solely on reviewer assignment. It explicitly avoids approving, merging, or commenting on pull requests, adhering to the principle of least privilege.

assign-prs