The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes local shell commands to facilitate environment setup and source control management. This includes git branch operations, remote repository synchronization, and worktree isolation for parallel development.
Evidence: SKILL.md contains logic for detecting default branches and managing feature branch transitions using git checkout and git pull commands.
[EXTERNAL_DOWNLOADS]: The skill connects to well-known external development and design services to fulfill its workflows. These interactions are performed through authenticated CLI tools and official MCP interfaces.
Evidence: references/tracker-defer.md details integration with Linear, Jira, and GitHub Issues. references/personas/ce-figma-design-sync.md facilitates capturing design specifications from Figma URLs.
[PROMPT_INJECTION]: The skill has an indirect prompt injection surface when it processes untrusted research artifacts, documents, or transcripts provided by the user.
Evidence: references/non-code-execution.md (Step 2) instructs the agent to read sources like PDFs and transcripts to synthesize deliverables without explicitly requiring the use of boundary markers or output sanitization. Ingestion points: references/non-code-execution.md. Capability inventory: SKILL.md (shell execution, file modification, and subagent dispatch). Sanitization: Absent.

ce-work