The Agent Skills Directory

[Indirect Prompt Injection] (HIGH): The skill allows the agent to process potentially untrusted input to generate migration logic and file names, creating a high-risk capability-to-content bridge.
- Ingestion points: User-provided migration names and SQL content (SKILL.md, Steps 1 and 2).
- Boundary markers: Absent; there are no delimiters or explicit instructions to ignore embedded commands in user-provided data.
- Capability inventory: Execution of shell commands via task and docker, writing to the local file system (Go files), and executing arbitrary SQL via tx.ExecContext.
- Sanitization: Absent; no validation or escaping mechanisms are described for the migration names or SQL strings.
[Dynamic Execution] (HIGH): The skill facilitates the creation of Go files in the migrations/ directory that are subsequently compiled and executed using task migrate:up. This runtime compilation and execution of generated source code is a high-severity pattern.
[Unverifiable Dependencies] (LOW): The skill references github.com/pressly/goose/v3. As this is a well-known and established library in the Go ecosystem, it is treated as a trusted source under [TRUST-SCOPE-RULE].

create-migration