The Agent Skills Directory

[COMMAND_EXECUTION]: The skill is designed to execute shell operations, run test suites, and interact with system toolchains as part of its primary function.
Evidence: SKILL.md explicitly lists shell, process, and test実行 (test execution) as handled components.
Evidence: integrated-workflow.md includes YAML schemas for planned_commands used during platform-specific validation steps.
[DATA_EXFILTRATION]: The skill reads existing source code, diffs, and project requirements from the local workspace to perform its analysis and implementation tasks.
Evidence: Phase 8 and Phase 9 of the integrated workflow involve reading and validating repository content.
[PROMPT_INJECTION]: The skill processes external, potentially untrusted data such as code diffs and requirement catalogs, creating a surface for indirect prompt injection.
Ingestion points: Workspace files, code diffs, and user-provided requirements (Phase 2).
Boundary markers: Use of highly structured YAML schemas and explicit phase transitions to separate data processing from execution.
Capability inventory: Shell command execution, filesystem read/write access, and toolchain invocation.
Sanitization: Implementation of "Hard Gates" and an independent verification phase (Phase 9) to ensure that generated actions match defined requirements and safety constraints.

mino-reproducible-development