The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes work descriptions from external specs or tickets, which is an inherent data ingestion surface.
Ingestion points: Work descriptions in specifications or tickets mentioned in SKILL.md.
Boundary markers: No explicit delimiters are specified for separating user data from system instructions.
Capability inventory: Execution of typechecking tools, individual test files, and full test suites (SKILL.md).
Sanitization: Input data sanitization is not mentioned in the skill instructions.
[SAFE]: No malicious patterns, obfuscation, or unauthorized access attempts were detected. The skill utilizes restrictive platform configurations, including disable-model-invocation: true and allow_implicit_invocation: false, to minimize potential risk vectors.

implement