The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to use the system date command to generate phase timestamps and provides example commands like cargo test and cargo build for inclusion in generated task files. These are benign and standard for a software development assistant.
[INDIRECT_PROMPT_INJECTION]: The skill possesses an indirect injection surface as it processes external, potentially untrusted project data (e.g., files in spec/ or .claude/ directories) to generate its plans.
Ingestion points: Files located in spec/*.md and .claude/skills/control-tower/rules/.
Boundary markers: None identified beyond standard Markdown formatting.
Capability inventory: It invokes other skills (test-design-review, subagent-driven-development) and generates executable scripts/tasks for subsequent agent actions.
Sanitization: No specific sanitization or filtering of input data is performed during prompt interpolation.
[REMOTE_CODE_EXECUTION]: There is no evidence of remote code execution or unauthorized package installation. The skill relies on local project context and explicitly mandates a user approval 'Hard-Gate' before moving from planning to implementation.

planning