A technical program manager maintains a docs-quality agent that must apply the team's approved "release-notes" skill on every sprint-close issue plus an Actions automation that generates changelogs nightly. The team wants consistent invocation without copying long prompts into each run. Which pattern should the manager use to call skills from agents and workflows effectively?
Select an answer to reveal the explanation.
Short Explanation
You want a speed-dial button, not a speech you recite each time. Referencing the skill by name lets agents pull the right playbook on demand. Once named, consistency follows everywhere it runs.
Full Explanation
Skills are designed to be invoked by name or path from agent definitions, custom agent instructions, and workflow prompts, allowing the runtime to discover and load the versioned instructions and resources consistently without duplication. Pasting full instructions into every issue comment satisfies execution once but misses the maintainability requirement because updates require editing history and risk drift across runs. Storing text in a separate wiki page satisfies documentation but misses automatic loading because agents and workflows do not reliably resolve ad hoc links as executable skills. Configuring an MCP connection satisfies tool and data access but misses procedural knowledge because MCP provides context and actions, not the named reusable instructions that define how to apply release-notes style. For the exam, remember invocation is by reference, not by copy, and in practice verify the skill name resolves in both interactive agents and automation runs after renaming.