The Agent Skills Directory

[DYNAMIC_CONTEXT_INJECTION]: The skill employs the !command syntax to execute a bash script (yaml-helper.sh) at skill load time. This mechanism is used to resolve configuration keys for automation and workflow modes. Execution of shell commands during the initial loading phase is a security concern as it occurs before AI safety review.\n- [COMMAND_EXECUTION]: The skill uses the Bash tool to run local scripts such as review-receipts.sh and adr-dep-graph.sh. These scripts are responsible for checking file hashes and calculating dependency orders for architectural decisions. These operations involve executing code within the project environment.\n- [INDIRECT_PROMPT_INJECTION]: The skill analyzes external design data, creating a surface for indirect instructions. 1. Ingestion points: The skill reads content from all Game Design Documents (design/gdd/*.md) and Architectural Decision Records (docs/architecture/adr-*.md). 2. Boundary markers: No explicit boundary markers or 'ignore' instructions are present for the ingested content. 3. Capability inventory: The skill has the capability to write files (Write), execute local bash scripts (Bash), and spawn further sub-agents (Agent). 4. Sanitization: The skill uses targeted Grep extraction to focus on technical sections, which reduces the data volume but does not perform sanitization of the extracted content.

architecture-review