The Agent Skills Directory

[SAFE]: The skill is restricted to documentation management tasks, specifically reading and writing markdown files in a designated documentation directory. It does not perform network operations, access sensitive system paths, or execute external scripts.
[SAFE]: The skill implements a controlled workflow for initializing documentation directories, requiring the target path to be empty or non-existent, which prevents accidental modification or overwriting of existing project files.
[INDIRECT_PROMPT_INJECTION]: The skill ingests data from repository files like README.md and AGENTS.md to generate specification context. The impact is assessed as safe because the agent is constrained by specific markdown templates and instructed to limit interaction to documentation drafting within a local environment.
Ingestion points: Project documentation files (e.g., README.md, existing spec files) and user input.
Boundary markers: Structured markdown templates and specific YAML frontmatter requirements.
Capability inventory: Reading and writing local repository markdown files.
Sanitization: Relies on structural constraints and template-driven output rather than explicit content filtering.

building-product-specifications