The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill facilitates the processing and generation of deployment plans, creating a surface where malicious instructions embedded in external files or metadata could influence the agent's behavior.
Ingestion points: The scripts/validate_deploy.py script reads external content from files (e.g., deploy-plan.md); scripts/generate_deploy.py incorporates user-supplied command-line arguments into generated artifacts.
Boundary markers: While SKILL.md defines a "Permission Boundary" requiring user approval for mutations, there are no technical delimiters or "ignore instructions" prompts used when interpolating or reading external data.
Capability inventory: The skill is granted Bash tool access along with file manipulation permissions (Read, Write, Edit, Glob).
Sanitization: Neither the validation nor the generation script performs sanitization or escaping of the ingested data.

deployment-engineer