The Agent Skills Directory

[SAFE]: The skill incorporates significant security guardrails, such as the OMA execution policy which explicitly prevents the agent from performing software builds or packaging until authorized by the user. Additionally, the common checklist in references/_shared/core/common-checklist.md requires that no secrets are present in code or logs and that all untrusted input is validated.
[INDIRECT_PROMPT_INJECTION]: The skill has an inherent attack surface as it processes user requirements and integrates with existing codebases, but it manages this through structured protocols.
Ingestion points: UI requirements, component specifications, and existing route data are ingested during the entry and 'ACQUIRE' phases as described in SKILL.md.
Boundary markers: The skill enforces a strict structural flow (Prepare, Acquire, Act, Verify) and uses dedicated markdown sections to isolate instructions from data.
Capability inventory: The agent can modify files within the CODEBASE scope and execute local processes for linting, type-checking, and testing (PROCESS scope) as defined in SKILL.md.
Sanitization: The skill mandates the use of Zod for data validation and follows framework-specific security patterns for React and Angular to mitigate injection risks.

oma-frontend