The Agent Skills Directory

[COMMAND_EXECUTION]: The skill's core functionality is to execute arbitrary shell commands across multiple local repositories as defined in a user-provided YAML specification file. The transform.command and verify[].command fields in the spec are executed using the Bash and Agent tools. While this is the intended purpose of the migration tool, it grants the agent broad execution capabilities over the local filesystem and git repositories.
[INDIRECT_PROMPT_INJECTION]: The skill represents an attack surface for indirect prompt injection because it ingests and acts upon instructions (commands) provided in swarm-specs/**/*.yaml files.
Ingestion points: Specification files located in swarm-specs/.
Boundary markers: The skill does not implement specific boundary markers or 'ignore' instructions for the content within the YAML command strings.
Capability inventory: The skill utilizes Bash, Agent, Write, Edit, and gh (GitHub CLI) tools to carry out the migration tasks.
Sanitization: There is no evidence of sanitization or restricted execution for the commands provided in the YAML spec, although the skill performs a 'Phase 1' validation for existence and basic syntax.
[DYNAMIC_EXECUTION]: The tool dynamically assembles and executes shell scripts at runtime based on the logic defined in the external YAML configuration, involving parallel subagents to perform transformations and verifications.

swarm-migrate