Setup-My-Iq

# Setup My IQ: Create and Update Context Portfolio

Create and maintain the personal context portfolio: a set of structured markdown

files that represent who you are, how you work, and what matters to you. The files

can live anywhere on disk (OneDrive folder, a git repo, a local directory, etc.).

What ties them into the agent is a pointer in one of the user's custom instructions

files (`AGENTS.md`, `copilot-instructions.md`, `CLAUDE.md`, or any `*.instructions.md`

that the host loads into the session).

How This Skill Works

When invoked, determine the current state and take the right action.

Discovering existing context

Scan the session for references to the files this skill creates. The host has

already merged all relevant custom instructions files into the session:

`AGENTS.md` (workspace or user-scope), `copilot-instructions.md`,

`CLAUDE.md`, any `*.instructions.md` file, and similar host-specific equivalents.

It doesn't matter which file the pointer came from, just look for the filenames:

- `identity.md`

- `role-and-responsibilities.md`

- `team.md`

- `tools-systems-and-config.md`

- `communication-style.md`

- `preferences-and-constraints.md`

For each, find an `@<absolute-path>` (or equivalent path reference) anywhere in

the loaded instruction text. That's the file's location. Don't rely on the label

or variable name next to the path: users may name pointers differently

(`identityProfile`, `me`, `who-i-am`, etc.). Match on the filename at the end

of the path.

For each filename, classify as:

- **NOT REFERENCED** — no path to this filename is loaded in the session.

- **REFERENCED, FILE MISSING** — a path is loaded but the file doesn't exist on disk.

- **PRESENT** — path is loaded and the file exists. Read it to check for placeholders.

If the six filenames are in mixed states, handle them in this priority order:

first create any NOT REFERENCED files via the interview, then fill REFERENCED,

FILE MISSING entries, then fix incomplete fields in PRESENT files. Summarize

the combined state to the user before starting so they know what to expect.

Concrete example. Suppose `identity.md` and `team.md` are PRESENT and

complete, `tools-systems-and-config.md` is PRESENT but has TODO placeholders

in the ADO section, `role-and-responsibilities.md` and `communication-style.md`

are REFERENCED but the files don't exist on disk, and

`preferences-and-constraints.md` is NOT REFERENCED. The skill should tell the

user: "You have identity and team done. Tools-config has gaps. Role and

communication-style are referenced but missing. Preferences isn't set up at

all. I'll interview you for preferences first, then create role and

communication-style, then fill the gaps in tools-config. OK?"

1. No context files referenced

None of the expected filenames appear anywhere in the loaded instructions.

Treat this as first-time setup. Run **First-Time Setup** below.

2. Some files referenced but missing on disk

The user started setup previously and didn't finish, or paths exist without

files behind them. **Resume where they left off:**

- List which files exist and which are missing.

- Tell the user: "You have {existing files}. Still need: {missing files}. Want to

pick up where you left off?"

- When choosing the next file to interview for, the priority order from

*Discovering existing context* still applies. Within a tie, use this

default sequence: identity -> role -> team -> tools -> communication-style

-> preferences-and-constraints.

3. Files exist but have incomplete fields

Files exist but some contain unfilled placeholder values. **Fill gaps:**

- Scan all referenced files for any of these incomplete-field patterns:

- `<!-- TODO -->` — explicit placeholder

- HTML comments of the form `<!-- ... -->` used as stand-in values

(e.g., `<!-- your name -->`, `<!-- org name -->`, `<!-- manager name -->`)

- Any table cell or field whose only content is an HTML comment

- List the gaps: "I found incomplete fields in {files}.