The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to utilize repository-specific tooling, specifically Python scripts located in .claude/scripts/code_graph. These tools are used for static analysis, such as tracing code dependencies, blast radius, and caller/callee relationships during the implementation and investigation phases.
[PROMPT_INJECTION]: The instructions contain directives to prioritize the skill's specific protocol over platform-level mode-switch instructions and mandate strict adherence to the defined task order. This is a common pattern in complex AI agent workflows to prevent context drift and ensure procedural compliance.
[DATA_EXPOSURE]: The skill requires the agent to read internal project configuration and documentation files (e.g., docs/project-config.json, docs/project-reference/) to ground its actions in the project's specific conventions. This is standard behavior for repository-aware coding assistants.
[EXTERNAL_DOWNLOADS]: The skill references setup routines like $project-init or $sync-codex to be executed if project configuration is missing or stale. These are internal project commands and do not point to untrusted external sources.

feature-implement