The Agent Skills Directory

Indirect Prompt Injection Surface: The skill is designed to ingest and process external design documents, requirements, and research data to generate decision records. This represents a potential surface where untrusted input could influence the agent's behavior.
Ingestion Points: User-provided requirement descriptions and existing documents in the docs/decisions/ directory.
Boundary Markers: While the skill enforces a specific Markdown structure for its output, it does not explicitly define delimiters or 'ignore' instructions for the data it reads.
Capability Inventory: The skill involves executing build commands (such as MSBuild), running tests with specific filters, and interacting with PR media tools (github-pr-media).
Sanitization: The skill proactively mitigates data exposure by instructing the redaction of secrets, personal data, and credentials before recording evidence or publishing artifacts.
Controlled Command Execution: The skill facilitates the execution of tests and build tools. While these are powerful capabilities, the skill provides specific guidance on scoping these actions, such as using isolated task-owned comparison worktrees and repository-pinned toolchains to ensure environment integrity.

feature-delivery