The Agent Skills Directory

[COMMAND_EXECUTION]: The skill directs the agent to execute several local utility scripts and Git commands to manage the implementation process. Examples include bash scripts/get-config, node scripts/spec-superflow.mjs, git worktree, and custom scripts like scripts/task-brief. These operations are used for configuration retrieval, task extraction, and project state management.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface as it processes external project data and passes it to subagents.
Ingestion points: Project files such as execution-contract.md, tasks.md, design.md, and various specification files in the specs/ directory.
Boundary markers: The subagent dispatch templates in implementer-prompt.md and task-reviewer-prompt.md use standard markdown headers (e.g., ## Task Description, ## Context) to separate inputs, but they lack explicit delimiters or warnings to ignore instructions embedded within those files.
Capability inventory: The skill has access to shell execution (bash), JavaScript execution (node), Git operations, and the ability to dispatch subagents with specific model configurations.
Sanitization: There is no evidence of sanitization, validation, or escaping of the content ingested from the project files before it is interpolated into subagent prompts.

execution-governor