The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the bash tool to execute a series of local Python scripts (feature-state.py, adr-query.py, learning-db.py) located in the agent's home directory (~/.claude/scripts/). These scripts are responsible for initializing feature states, reading hierarchical context, and registering architectural decisions. This behavior is consistent with the skill's intended purpose for internal workflow management.
[PROMPT_INJECTION]: The skill presents an indirect prompt injection surface by interpolating user-provided feature names and descriptions into shell command arguments. \n
Ingestion points: User-provided feature description and triggered feature name inputs. \n
Boundary markers: The instructions enclose arguments in double quotes (e.g., "FEATURE_NAME") to prevent simple word splitting. \n
Capability inventory: The skill is allowed to use Bash, Write, Read, Grep, and Edit tools. \n
Sanitization: There is no explicit instruction to sanitize or escape shell metacharacters (such as backticks or subshells) from the user input before execution.

feature-design