The Agent Skills Directory

[COMMAND_EXECUTION]: The skill is designed to interact with the PlanetScale CLI (pscale). It instructs the agent to execute various commands for database management, configuration, and inspection.
Evidence: Usage of pscale branch changes, pscale branch data-topology, pscale branch shard, pscale branch config-profile, and pscale role get throughout SKILL.md.
Safety Measures: The skill mandates using --format json for reliable parsing, explicit --org flags, and user approval for all destructive or cost-incurring mutations.
[INDIRECT_PROMPT_INJECTION]: A potential vulnerability surface exists where the agent is instructed to update the branch data topology from a local JSON file (topology.json). If the content of this file is untrusted, it could influence the agent's context.
Evidence: pscale branch data-topology update <database> <branch> --org <org> --format json < topology.json in SKILL.md.
Mitigation: The instructions explicitly state that "Empty input, arrays, scalars, and malformed JSON are rejected" and require the user to "Review the complete JSON object... before applying it."
[SAFE]: The skill demonstrates defensive security practices by providing checksums for the CLI binaries used to generate the documentation, ensuring that the commands described are from verified versions.
Evidence: Reference to SHA-256 61eadf71c9d5e6423587fbf01fd698800d8a944775a06519a1c2ac38fcb89287 and others in references/commands.md.

pscale-neki