SKILL_NAME

{{SKILL_TITLE}}

{{ONE_PARAGRAPH_ON_WHAT_THIS_SKILL_DOES_AND_WHAT_IT_HANDS_BACK}}.

SCAFFOLD NOTE — delete this paragraph and every other one marked the same way; they are addressed to you rather than to the agent that will read this skill, and the review step fails while any remain. Rename this directory and the name: above together: a skill's name: must equal its directory name, and nothing at runtime reports a mismatch.

Underneath, memory-search owns the search ladder and the relevance gate, graph-traversal owns relationship walking, and memory-capture owns any durable write. This skill adds only what is specific to {{THE_ROLE_ACTIVITY}} — the inputs to read, the order to read them in, and the shape of the output. Do not restate their rules here. A second copy of the search ladder is a copy that drifts, and the reader cannot tell which one is current.

Hard rules (non-negotiable — read first)

1. The ticket tracker and its wiki are the system of record. Read the ticket and its linked pages before memory, and report against them. Never propose tracker mechanics this organization does not use — the label, template, and milestone conventions a public repository runs on do not transfer to closed delivery work.
2. Produce input, never a committed number or a written field. {{WHAT_THIS_SKILL_OUTPUTS}} is a proposal for the team to argue with. Write nothing into the tracker, and post no comment, without approval of the exact text in the session.
3. Cite every claim that rests on a record. A finding drawn from memory or from a page quotes and names its source. An uncited claim presented as the team's standard cannot be argued with by the person it is aimed at.
4. Engagement scope on every write. Pass the group explicitly. Client work does not cross tenants: nothing learned on one engagement is written into another's scope, and nothing client-identifying goes into a shared or personal scope.
5. Say what was not found. An empty recall is a result. Report it rather than filling the gap with a plausible-sounding inference.