メニュー
Use when the user asks to create, draft, save, revise, or improve a Codex Goal, GOAL.md, goal file, persistent objective, completion contract, or long-running Codex work objective. This skill must be used before inventing or saving any goal for Codex.
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
Multi-phase, subagent-driven deep audit and remediation of codebase guidance (root agent instructions, docs, module context files, skills/agents references, task routing). Phase 1 subagents write verified findings only; Phase 2 subagents verify and fix guidance. Use when the user asks to audit, refresh, or fix agent context, CLAUDE.md or AGENTS.md drift, stale docs, contradictory rules, or run a context-guidance deep dive.
Use when the user wants to create, brainstorm, draft, or post social media content for X (Twitter), LinkedIn, or Instagram. This includes generating post ideas, writing captions or threads, adapting content across platforms, planning a posting cadence, or navigating to a platform's posting interface via browser tools. Trigger whenever the user mentions social media, posting, tweets, LinkedIn posts, Instagram captions, content calendars, thought leadership posts, or sharing research/business updates online — even if they just say something like 'I should post about this' or 'help me share this.'
Build Discounted Cash Flow (DCF) valuation models for commercial real estate (CRE). Calculate NOI-based cash flows, levered and unlevered IRR, equity multiple, DSCR, debt yield, exit cap rate reversion, and sensitivity analysis. Use for office, retail, industrial, multifamily, mixed-use, and hotel properties. Trigger on: CRE valuation, property DCF, NOI projection, cap rate analysis, IRR analysis, real estate investment return, acquisition underwriting, hold-period analysis, CRE sensitivity table.
Build discounted cash flow (DCF) valuation models in Excel specifically for commercial real estate (CRE). Use when creating CRE acquisition underwriting models, property-level DCF workbooks, NOI projections, rent roll buildouts, debt schedules, or IRR/equity-multiple return analyses in Excel or ExcelJS. Covers all CRE property types: office, retail, industrial, multifamily, mixed-use, hotel. Trigger on: 'CRE excel model', 'build underwriting model', 'property DCF excel', 'NOI projection spreadsheet', 'rent roll model', 'IRR excel', 'acquisition model', 'CRE waterfall', 'cap rate model'.
Create new skills, modify and improve existing skills, and measure skill performance. Use when users want to create a skill from scratch, update or optimize an existing skill, run evals to test a skill, benchmark skill performance with variance analysis, or optimize a skill's description for better triggering accuracy.
Use when auditing, optimizing, or architecting the AI agent context layer (CLAUDE.md files, hooks, slash commands, skills, IDE rules) for any codebase, bootstrapping context engineering from scratch, diagnosing agent underperformance, or when the user mentions CLAUDE.md strategy, context quality, agent instructions, or context architecture.
| name | goal-writer |
| description | Use when the user asks to create, draft, save, revise, or improve a Codex Goal, GOAL.md, goal file, persistent objective, completion contract, or long-running Codex work objective. This skill must be used before inventing or saving any goal for Codex. |
Write concise Codex Goal files using the OpenAI Goals pattern: a goal is a scoped completion contract, not a vague task prompt.
Save one Markdown file under goals/ named:
YYYY-MM-DD-<short-slug>-goal.md
The saved goal body must be about 3000-3500 characters, excluding frontmatter if any. Do not use GOAL.md, goal.md, or undated names.
Use these headings:
# Codex Goal - <short title>
## Goal
## Boundaries
## Iteration Policy
## Verification
## Deliverables
## Blocked Stop Condition
goals/YYYY-MM-DD-<short-slug>-goal.md./goal.Do not mark a goal complete just because a budget is reached. Budget exhaustion means summarize progress, blockers, and the next useful step.
Use a short lowercase slug from the objective:
2026-05-20-citation-refresh-hardening-goal.md2026-05-20-prospecting-browser-qa-goal.md2026-05-20-replication-evidence-audit-goal.mdThis follows OpenAI's "Using Goals in Codex" guidance: strong goals name the outcome, verification method, and constraints, and research goals must separate confirmed evidence, support-only evidence, blockers, and remaining uncertainty.