The Agent Skills Directory

[PROMPT_INJECTION] (LOW): Indirect Prompt Injection surface identified.
Ingestion points: The skill reads untrusted data from design.md and prd.md within the .claude/cx/features/ directory.
Boundary markers: There are no explicit boundary markers or 'ignore' instructions provided when interpolating external design content into the new task files.
Capability inventory: The skill has the capability to write files (task-{n}.md, status.json) which are then consumed by subsequent execution agents (e.g., /cx-exec).
Sanitization: No visible sanitization or validation of the text extracted from design documents before it is used to generate task descriptions.
[COMMAND_EXECUTION] (SAFE): Use of standard local utilities for environment setup.
Evidence: The skill uses git rev-parse --show-toplevel, jq, and mkdir -p to initialize the project directory and read configuration.
Context: These commands are used for legitimate configuration management within the local developer environment and do not involve executing external or untrusted code.

cx-plan