The Agent Skills Directory

Dynamic Adapter Dispatching: The skill identifies and executes adapter-specific logic based on entries in the project's configuration. It constructs file paths to these adapters (e.g., tools/forwarder-relay/<name>/) at runtime. While this enables extensibility, it involves dynamic loading from computed paths. This is a core part of the skill's primary function and is mitigated by the requirement that adapters reside within a specific, predefined directory structure.
Processing of Untrusted External Content: As a triage tool for security reports, the skill ingests email headers and bodies which originate from external sources. This presents an indirect prompt injection surface. The skill includes explicit instructions to treat all message content as data rather than instructions and relies on a pre-flight redaction step performed by parent skills to minimize risks.
Local Pre-flight Command Execution: The skill performs an initial setup check using a local Python module (setup_preflight). This execution is used to verify the integrity of the skill's configuration and environment before processing any data, which is a standard verification practice in this framework.
Confidentiality and Mutation Controls: The skill strictly follows a 'propose, never apply' model, ensuring that no state changes—such as creating tracker issues or sending emails—occur without explicit human review. It also enforces strict confidentiality rules for all data extracted from private security reports.

issue-import-via-forwarder