The Agent Skills Directory

[COMMAND_EXECUTION]: The skill explicitly uses the Bash tool to interact with the environment. It executes unity status, project compilation commands, and NUnit test runners. While these are necessary for its stated purpose, the use of shell execution represents an active attack surface.
[DYNAMIC_EXECUTION]: The workflow requires compiling Unity projects and executing EditMode and PlayMode tests. This involves the runtime execution of code defined within the repository, which is a form of dynamic execution triggered by the agent.
[INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it processes untrusted data from the repository.
Ingestion points: The agent reads project-local instructions (AGENTS.md, CLAUDE.md), task contracts (production/tasks/<TASK-ID>/contract.md), and Unity console logs (references/claude-skill.md).
Boundary markers: No explicit boundary markers or instructions to ignore embedded commands within the ingested files are defined.
Capability inventory: The skill has access to Read, Glob, Grep, Bash, and Write tools, allowing it to execute shell commands and modify the file system (e.g., writing reports to production/qa/).
Sanitization: There is no evidence of sanitization or filtering of the content read from the project files before it is processed by the agent.

smoke-check