The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests user prompts to compose design pipelines and delegates tasks to persistent workers and sub-agents.
Ingestion points: User prompts processed in SKILL.md and evals/evals.json to determine routing logic.
Boundary markers: The skill instructs the agent to use "DESIGN CONTEXT" and "SKILL" tags to wrap context when prompting workers in SKILL.md section "DEFER — cross-stage contracts".
Capability inventory: Shell execution via jq and head, persistent worker spawning via agent-tmux, and file writes to DESIGN.md across various scripts.
Sanitization: No explicit instructions for sanitizing or escaping user-supplied strings before they are interpolated into delegated worker prompts.
[COMMAND_EXECUTION]: The skill utilizes shell commands to inspect its environment and configuration.
Executes jq -r '.skills|keys[]' ~/.agents/.skill-lock.json to retrieve the roster of installed skills.
Executes head -8 on member skill files to check for invocation constraints (disable-model-invocation).
[EXTERNAL_DOWNLOADS]: The skill references an external web resource in references/design-roles.md.
Provides a link to https://namethatui.com/ as a dictionary reference for UI component naming conventions.

using-design-skills