The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill provides a feature to analyze existing codebases via the /ace:map-codebase command. This ingests untrusted data from local project files into the agent's context, creating an indirect prompt injection surface. A malicious codebase could attempt to influence the orchestrator's planning or sub-agent instructions. However, this is a known risk for any code-analysis tool and is mitigated by the skill's structured Markdown templates and planning process.
[COMMAND_EXECUTION]: The skill relies on the Bash tool to perform essential project lifecycle operations, including initializing Git repositories, committing changes atomically per task, and running build or linting commands. These operations are transparently defined in the command files and follow standard development practices.
[REMOTE_CODE_EXECUTION]: The system architecture uses the Task tool to spawn parallel Team Lead agents for executing plans. While this allows for autonomous code implementation by sub-agents, it is the fundamental design of the orchestration system and is governed by strict zone-based ownership rules defined in project-types.json to prevent unintended file modifications.

ace