How do I install the REPLACE-skill-name skill?

Run npx skillmds@latest add aircfo/replace-skill-name in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

Is REPLACE-skill-name safe to use?

This skill has not completed SkillMD's automated safety review yet. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with REPLACE-skill-name?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is REPLACE-skill-name free to use?

Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.

Who published REPLACE-skill-name?

aircfo (@aircfo) published this skill. Their other Agent Skills are listed on their SkillMD profile.

REPLACE-skill-name by aircfo — Agent Skill

Name: REPLACE-skill-name
Author: aircfo

REPLACE-skill-name

REPLACE: what this skill produces, then the phrases that should trigger it. Pattern to follow: <One sentence on the outcome, naming the systems it reads.> Use when someone says "REPLACE phrase one," "REPLACE phrase two," "REPLACE phrase three," or is doing REPLACE-the-task-context. This description is the trigger surface — an agent decides whether to run the skill from this text alone, so be specific about both the outcome and the wording a real user would use.

aircfo 3537aa0 2.7 KB Updated Sep 17, 2026

File contents

REPLACE: Skill Title — what it produces

REPLACE: one short paragraph naming the deliverable and who reads it.

What this produces

REPLACE: describe the concrete output — a file, a table in chat, a workbook, a drafted summary. Be specific about where it's written and what it's named.

Principles

REPLACE: the non-negotiables for this skill. Keep the ones that apply:

Read-only. Never create, modify, send, or delete anything in a financial system.
Evidence or silence. Every number cites where it came from. Never present an estimate as a figure from the books.
Fill or placeholder — never guess. Mark gaps explicitly and surface them; don't invent a value to complete a section.
Confirm before anything outward-facing. Draft; never send.
Degrade gracefully on partial access. A blocked endpoint is partial, not failed: read what you can and mark the rest.
Always pass explicit date ranges on dated report pulls — both a start and an end date.

Step 0 · Load the ContextOS

REPLACE or delete. If this skill reasons about company-specific finance definitions, load the finance profile first so it uses the company's own definitions of burn, runway, and revenue:

Look for the pointer file .aircfo-finance-context.json in the workspace; fall back to ~/Desktop/ContextOS/finance-profile.md, then to the legacy ~/Desktop/airCFO Finance Context/finance-profile.md. If no profile exists, say so and offer to run /finance-context-builder first — then continue with stated assumptions rather than silently inventing definitions.

Steps

1 · REPLACE: first step

REPLACE: what to do, which tools to call, what to show the user.

2 · REPLACE: second step

REPLACE.

3 · REPLACE: present the result

REPLACE: the exact output shape. If there's a review gate, say what the user is ruling on and wait for their answer before continuing.

Output format

REPLACE: show the literal shape of the deliverable — a sample table, file tree, or section headings — so runs are consistent.

Failure modes

REPLACE: what to do when a connector isn't authorized, returns no data, or returns numbers that look implausible. Never fabricate a result to complete the run; report what's missing.

---name: replace-skill-namedescription: REPLACE: what this skill produces, then the phrases that should trigger it. Pattern to follow: <One sentence on the outcome, naming the systems it reads.> Use when someone says "REPLACE phrase one," "REPLACE phrase two," "REPLACE phrase three," or is doing REPLACE-the-task-context. This description is the trigger surface — an agent decides whether to run the skill from this text alone, so be specific about both the outcome and the wording a real user would use.---# REPLACE: Skill Title — what it producesREPLACE: one short paragraph naming the deliverable and who reads it.## What this producesREPLACE: describe the concrete output — a file, a table in chat, a workbook, a draftedsummary. Be specific about where it's written and what it's named.## PrinciplesREPLACE: the non-negotiables for this skill. Keep the ones that apply:- **Read-only.** Never create, modify, send, or delete anything in a financial system.- **Evidence or silence.** Every number cites where it came from. Never present an estimate as a figure from the books.- **Fill or placeholder — never guess.** Mark gaps explicitly and surface them; don't invent a value to complete a section.- **Confirm before anything outward-facing.** Draft; never send.- **Degrade gracefully on partial access.** A blocked endpoint is partial, not failed: read what you can and mark the rest.- **Always pass explicit date ranges** on dated report pulls — both a start and an end date.## Step 0 · Load the ContextOSREPLACE or delete. If this skill reasons about company-specific finance definitions, loadthe finance profile first so it uses the company's own definitions of burn, runway, andrevenue:> Look for the pointer file `.aircfo-finance-context.json` in the workspace; fall back to> `~/Desktop/ContextOS/finance-profile.md`, then to the legacy> `~/Desktop/airCFO Finance Context/finance-profile.md`. If no profile exists, say so and> offer to run `/finance-context-builder` first — then continue with stated assumptions> rather than silently inventing definitions.## Steps### 1 · REPLACE: first stepREPLACE: what to do, which tools to call, what to show the user.### 2 · REPLACE: second stepREPLACE.### 3 · REPLACE: present the resultREPLACE: the exact output shape. If there's a review gate, say what the user is ruling onand wait for their answer before continuing.## Output formatREPLACE: show the literal shape of the deliverable — a sample table, file tree, or sectionheadings — so runs are consistent.## Failure modesREPLACE: what to do when a connector isn't authorized, returns no data, or returns numbersthat look implausible. Never fabricate a result to complete the run; report what's missing.

Frequently asked questions