About this skill

Problem

Role-agent prompts often mix capabilities, knowledge, workflows, and permissions in one long description, making them hard to mount, search, and maintain. This skill breaks the capabilities in a role-agent prompt into reusable Agent Skill directories, with traceable knowledge, source maps, and a checkable delivery structure.

How It Works

One capability, one directory: by default, each core role capability maps to one SKILL.md directory, without silent merging or splitting.
Traceable knowledge: it retrieves from Tencent IMA or local files, extracts references/knowledge.md, and records references/source-map.md; when evidence is weak, it writes knowledge-gaps.md.
Prompt write-back: it updates the capability layer, knowledge layer, and mapping table so capabilities, skill directories, and knowledge docs stay aligned while preserving the seven-layer structure.

Boundaries

It is best for prompts that already exist and need to be organized into a skill package. If the knowledge base is unavailable, evidence is thin, or capability granularity is unclear, it produces skeletons, gap lists, or split/merge proposals first, rather than inventing unsupported content. It also expects final delivery to include mapping, self-check results, and pending-source lists.

Screenshots

Use Cases

Split core capabilities in a role-agent prompt into `SKILL.md` directories.
Extract IMA or local docs into `knowledge.md` and `source-map.md`.
List evidence gaps in `knowledge-gaps.md` and pending source lists.
Write back capability and knowledge layers plus the mapping table.

Best For

Prompt engineers maintaining role agents, who need capability notes turned into mountable skill packages.
AI knowledge-base owners who need traceable sources before skills are used in production.
Solution consultants implementing enterprise agents, who need SOPs organized into spec-compliant skill directories.
Product managers overseeing multi-role agents, who need mapping tables and gap lists.