The Agent Skills Directory

[SAFE]: The skill is primarily instructional and provides a structured planning workflow for backend development. It includes explicit safety rules prohibiting the use of secrets in contracts and forbidding the AI from making unauthorized network calls.
[INDIRECT_PROMPT_INJECTION]: The skill identifies surfaces where external data, such as API specifications or access logs, might be processed. 1. Ingestion points: User requirements and existing repository artifacts (e.g., OpenAPI specs). 2. Boundary markers: Explicitly defined in section 9 of the service-contract reference, which instructs the agent to treat contract text as data only. 3. Capability inventory: The skill body restricts core capabilities, forbidding the use of network or database calls. 4. Sanitization: Includes instructions to validate references against an allowlist and to treat external content as non-executable data.
[DATA_EXPOSURE]: Safety guidelines within the skill explicitly warn against placing secrets, tokens, or connection strings in generated contracts or handoff text.

omh-backend