The Agent Skills Directory

[DYNAMIC_CONTEXT_INJECTION]: The skill uses the exclamation-backtick syntax (!bash ...) to execute yaml-helper.sh when the skill is loaded. This is used to dynamically inject configuration data (automation, workflow, and system overrides) into the prompt at load time.
[COMMAND_EXECUTION]: The skill executes several shell commands and local scripts as part of its core logic:
It calls git log in Phase 2 to determine which documents have been modified since the last review.
It executes a project-local script .claude/scripts/registry.sh in Phase 6 to validate changes made to the entity registry.
These executions are restricted by the allowed-tools configuration but still represent a substantial execution surface within the project environment.
[INDIRECT_PROMPT_INJECTION]: The skill scans and processes content from multiple external files, creating an attack surface for indirect prompt injection.
Ingestion points: Files matching design/gdd/*.md and the design/registry/entities.yaml file are read and parsed.
Boundary markers: The instructions do not define clear boundary markers or safety instructions to prevent the agent from obeying instructions that might be embedded in the design documents being scanned.
Capability inventory: The skill possesses significant capabilities, including the ability to Write to the filesystem, Edit existing files, and execute Bash commands.
Sanitization: There is no evidence of sanitization or escaping of the data extracted from the Game Design Documents (GDDs) before it is used to generate reports or perform comparisons.

consistency-check