The Agent Skills Directory

[SAFE]: The skill follows a process-oriented design where the primary agent acts as a coordinator rather than a coder. It mandates that a 'Tester' subagent handle all code writing, while the main agent focuses on drafting 'Lock briefs' (Why/What/How) for user approval.
[SAFE]: No network requests (curl, wget, fetch), filesystem modifications (write/chmod), or privilege escalation commands (sudo) were detected. The skill primarily consists of markdown-based instructions and templates.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection as it is designed to process and analyze external project files and codebases.
Ingestion points: Project files, public entry points, and shared documentation in ../pack-shared/ and ../setup-toolkit/.
Boundary markers: The skill implements a functional boundary by requiring the user to explicitly approve a 'Lock brief'—which defines the scope and logic of the test—before any code generation is allowed to proceed.
Capability inventory: This skill explicitly disables model invocation (disable-model-invocation: true) and lacks direct file-writing or subprocess capabilities, delegating those to a subagent.
Sanitization: Security is maintained through a mandatory manual review gate where the user must approve the testing plan (Why/What/How) before the 'Tester' subagent is invoked.

create-test