The Agent Skills Directory

[COMMAND_EXECUTION]: The skill performs local command execution to facilitate the CUBRID development workflow. This includes running just build for compilation checks and various git and GitHub CLI (gh) operations to manage code diffs and PR context. These operations are performed within the user's local environment and are restricted to standard development tasks.
[COMMAND_EXECUTION]: Sources a shared utility script from a sibling directory (../cubrid-common/scripts/cubrid-common.sh) to perform environment validation. This is a common pattern for modular skills within the same repository or suite to share logic for detecting source trees and build presets.
[EXTERNAL_DOWNLOADS]: Fetches external data from GitHub (PR diffs and metadata) and JIRA (ticket descriptions) to provide context for the implementation tasks. These interactions use official CLI tools or internal skill calls to well-known developer services.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface as it processes external data from JIRA tickets, GitHub PRs, and spec files which are then passed to sub-agents for implementation and review.
Ingestion points: Untrusted data enters the context from JIRA tickets, PR descriptions/diffs, and local spec files during the context-gathering phase (Step 3).
Boundary markers: The orchestrator provides this data to sub-agents as named context variables without the use of explicit delimiters or instructions to ignore embedded commands.
Capability inventory: The writer sub-agent can modify local files (Edit, Write), and the skill orchestrator automatically executes build commands (just build) on those modifications.
Sanitization: There is no explicit sanitization of the external content before it is interpolated into the prompts for the writer and reviewer sub-agents.

cubrid-grill-and-implement