The Agent Skills Directory

[DYNAMIC_CONTEXT_INJECTION]: The skill uses the ! syntax in SKILL.md to execute a shell command when the skill is loaded. It calls a yaml-helper.sh script located in a relative path (../../hooks/) to resolve configuration settings for workflow and automation modes.
[COMMAND_EXECUTION]: The skill uses the Bash tool to perform configuration resolution. Access is restricted in the frontmatter to a specific script path, though the script allows trailing arguments.
[INDIRECT_PROMPT_INJECTION]: The skill processes user-controlled data to generate audit reports.
Ingestion points: Reads files like design/gdd/*.md, design/registry/entities.yaml, and various engine asset files (.tscn, .unity, .umap, .json, .yaml).
Boundary markers: Uses Grep patterns and Markdown headers to locate data, but lacks specific delimiters to isolate potential instructions within the design documents.
Capability inventory: Includes file system access (Read, Glob, Grep, Write) and restricted command execution (Bash).
Sanitization: No explicit sanitization or filtering is performed on the data extracted from design files before it is used to generate reports or recommendations.

content-audit