The Agent Skills Directory

[SAFE]: The skill operates entirely within the project scope using standard tools (Read, Grep, Glob, Write). It does not perform any network operations or access sensitive system paths.
[SAFE]: The skill implements a human-in-the-loop pattern for all critical actions, such as updating the systems index or the requirement registry, ensuring the user maintains control over the project state.
[SAFE]: The use of the Task tool to consult engine specialists is a standard agentic pattern for domain-specific validation and does not involve the execution of untrusted code.
[INDIRECT_PROMPT_INJECTION]: The skill has a standard attack surface for documentation-processing tools as it ingests content from project files.
Ingestion points: Markdown files in design/gdd/, docs/architecture/, and production/epics/ are read completely during analysis.
Boundary markers: The skill does not use explicit delimiters or 'ignore embedded instructions' warnings when processing extracted technical requirements.
Capability inventory: The skill has the ability to write to documentation files, spawn subagents via the Task tool, and ask the user questions.
Sanitization: There is no explicit sanitization of technical requirements extracted from design documents before they are used to build reports or matrices. However, the risk is mitigated by the requirement for user approval for all file write operations.

architecture-review