The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes project requirements and architecture documentation, which creates a surface for indirect prompt injection if those documents contain adversarial instructions.
Ingestion points: Project documentation located in docs/02-requirements/ and docs/03-architecture/ are used as primary inputs for code implementation and task management.
Boundary markers: The skill enforces structural boundaries through the "Clean 4 Layers" architecture and strict Test-Driven Development (TDD) cycles, which separate business logic from external frameworks.
Capability inventory: The skill has the capability to write source code files, execute shell commands for testing suites (pytest, npm test, go test), and interact with the GitHub API via the gh CLI tool.
Sanitization: There are no explicit instructions or automated mechanisms defined within the skill to sanitize or validate the content of ingested documents before processing.
[EXTERNAL_DOWNLOADS]: The skill configuration for rendering C4 diagrams includes a remote fetch for a PlantUML library.
Evidence: The file templates/c4.puml includes the directive !include https://raw.githubusercontent.com/plantuml-stdlib/C4-PlantUML/master/C4_Context.puml.
Context: This is a standard and well-recognized community library for creating C4 model diagrams.
[COMMAND_EXECUTION]: The skill instructions and templates facilitate the execution of developer tools and shell commands to manage the software lifecycle.
Evidence: SKILL.md and templates/test-plan.md instruct the agent to run pytest, npm test, go test, and behave. templates/issue-tracker-github.md provides patterns for using the gh CLI for issue and PR management.

ooad-build