The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external design documents and specifications which could potentially contain hidden malicious instructions aimed at influencing the agent's behavior.
Ingestion points: The agent is instructed to read an "approved design", "accepted review brief", or "spec" to extract entities and fields (SKILL.md, Phase 1).
Boundary markers: The skill uses a <HARD-GATE> requiring a confirmed implementation plan before writing code or running commands, providing a human-in-the-loop checkpoint. However, there are no explicit instructions for the agent to ignore prompt-like text within the ingested documents.
Capability inventory: The skill modifies local Go files (Phase 3) and prompts the user to execute shell commands such as go generate, buf, and atlas (Phase 5).
Sanitization: No specific sanitization or filtering logic is mentioned for the content extracted from the external designs.
[COMMAND_EXECUTION]: The skill instructs the user to run several command-line tools as part of the implementation and verification workflow.
Evidence: Commands include go generate ./ent/..., buf generate, atlas migrate diff, and go test ./... (SKILL.md, Phase 5; references/implementation-rules.md, Section 11). These are standard development tools for the Go/Ent ecosystem and are appropriate for the skill's stated purpose.

ent-schema-implementer