메뉴
Auto-generate or update description, type, topics, aliases, and graph-backed relationship fields for a file.
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
SOC 직업 분류 기준
| name | enrich |
| description | Auto-generate or update description, type, topics, aliases, and graph-backed relationship fields for a file. |
Add or update frontmatter metadata for a file to make it discoverable through progressive disclosure.
When the user invokes /enrich [file] or /enrich (no argument):
rg -L --glob "*.md" --files-without-match "^description:" Tasks/ Context/Knowledge/ Context/Read the file fully
Check existing frontmatter — what fields are present vs missing
Read Context/Memory/graph.yaml if it exists before proposing any relational metadata so you can resolve names against canonical node ids and aliases.
Auto-generate missing fields:
description (~150 chars): A retrieval filter, not a summary. It answers "should I load this file?" Include what makes the content distinctive — mechanism, scope, or implication. Must differ meaningfully from the title.
type: One of: task, prd, knowledge, meeting-note, decision-record, progress-update, learning, track, slice
topics: 1-3 tags from the Topic Vocabulary:
multi-agent, canva-ai, customer-experience, leadership, product-craft, strategy, voice-agent, localization, retention, team-cn
maturity (knowledge files only): seedling | developing | evergreen
aliases (canonical person / project pages only): shorthand or alternate names already used elsewhere in the workspace. Prefer [] over guessing.
people / projects / channels (task and meeting-note files only): arrays of graph node ids from Context/Memory/graph.yaml
Present the proposed frontmatter to the user:
Proposed enrichment for [filename]:
- description: "[generated description]"
- type: [type]
- topics: [topics]
- aliases: [aliases if relevant]
- people: [people ids if relevant]
- projects: [project ids if relevant]
- channels: [channel ids if relevant]
If any names were unresolved, list them under Unresolved references: and ask whether to add graph nodes or leave them out for now.
Ask: "Look good, or should I adjust anything?"
On confirmation, update the file's frontmatter (preserve all existing fields and content)
people / projects / channels must use graph node ids only — never display namesIf the user says /enrich batch [folder]:
Prepare for a 1:1 — pull context from goals, meeting notes, Slack, and tasks to generate talking points
Build a focused plan for today based on goals, tasks, and recent activity. Reads from the weekly plan if it exists and updates it.
Check system health — metadata gaps, graph integrity, alias sync, and referential integrity.
Weekly review + compound learnings — review progress, surface blockers, distill insights into Memory
Cross-initiative alignment checker — surfaces misalignment, gaps, duplication, dependency risks, and synergy opportunities across active initiatives in the company and your groups roadmap. Complementary to /pulse (broad discovery); /align focuses on whether your initiatives are pulling in the same direction and have synergies with other company groups. Run ad-hoc or before key decisions.
Autonomously optimize any Agent skill by running it repeatedly, scoring outputs against binary evals, mutating the prompt, and keeping improvements. Based on Karpathy's autoresearch methodology. Use when: optimize this skill, improve this skill, run autoeval on, make this skill better, self-improve skill, benchmark skill, eval my skill, run evals on. Outputs: an improved SKILL.md, a results log, and a changelog of every mutation tried.