The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the Bash tool to perform a wide array of local system operations, including executing engine-specific binaries (Godot, Unity, Unreal), running build processes, and executing npm test scripts. It also uses utility commands like timeout, gtimeout, and perl for process management.
[DYNAMIC_CONTEXT_INJECTION]: The SKILL.md file contains a dynamic context injection pattern (!bash ...) that executes a local helper script (yaml-helper.sh) at skill load time to resolve configuration keys such as qa.level and testing.strict into the active session context.
[INDIRECT_PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by ingesting and acting upon untrusted data from multiple project files.
Ingestion points: Loads data from project.yaml, production/sprints/ files, production/qa/smoke-tests.md, and production/qa/qa-plan-*.md.
Boundary markers: The skill expects structured data (YAML, Markdown tables) but does not utilize explicit delimiters or 'ignore' instructions to isolate ingested content from the agent's logic.
Capability inventory: The agent has broad shell execution capabilities (Bash) and file writing capabilities (Write) which could be exploited if malicious instructions were successfully injected through the ingested files.
Sanitization: There is no evidence of sanitization or validation performed on the strings retrieved from external project files before they influence the selection of test paths or coverage analysis.

smoke-check