# Prompt 2. Tailoring a CV without inventing anything

**Use it when:** you have decided to apply, and before you run `tools/cv-audit.py`.

**What it is for:** using their words for the things you genuinely did. Nothing
else. The whole value of this prompt is in what it refuses to do.

---

## The one rule

**No invention.** If you do not have the experience, no experience grows on the
page. The model may borrow the posting's own phrasing **only where you genuinely
have the skill being described**.

That rule sounds obvious and it fails quietly. The dangerous edits are not lies.
They are small, comfortable improvements that turn into lies two interviews
later, when somebody asks a second question and you hear yourself hedging.

---

## The prompt

```
You are tailoring my CV to one job posting. You have one rule that outranks
every other instruction here: NO INVENTION.

MY CV (the true one)
<paste it>

THE POSTING
<paste it>

WHAT YOU MAY DO

1. Reorder. Put the bullets that match their requirements higher.
2. Re-word, ONLY where I already have the skill: if they call it X and I called
   it Y, and X and Y are the same thing I actually did, use X.
3. Cut. Remove bullets that carry nothing for this role.

WHAT YOU MAY NOT DO

- Add a skill, tool, metric, scale, duration or responsibility that is not in my
  CV. Not "implied by". Not "reasonable given the seniority". Not there means
  not there.
- Upgrade a verb. "read about" is not "used". "used once" is not "built".
  "helped with" is not "owned". If my CV does not name the level, ask me.
- Round a number up, or make a vague number specific.
- Turn a team I was in into a team I led.

OUTPUT, in three parts:

PART 1: the tailored CV.

PART 2: a change log. One line per change, in this format:
  [REORDER] moved X above Y because the posting lists it as required
  [REWORD]  "my phrase" -> "their phrase"  (same thing, because ...)
  [CUT]     removed X because ...

PART 3: the gaps. List every requirement in the posting that my CV does NOT
support. Do not soften this list and do not suggest wording for it. For each
one say which it is:
  - hours       I could form an honest opinion in an afternoon
  - a weekend   I could run it for real and describe its shape
  - real work   I would have to actually do the job to claim this

Then, for every claim that survives in the tailored CV, write the follow-up
question an interviewer would ask about it. If I cannot answer one of those
questions out loud, that claim comes back out.
```

---

## After it runs

1. **Read PART 3 first.** The gap list is the useful output. PART 1 is just
   formatting. The gaps tell you whether to apply, and what to spend a weekend on.
2. **Read every `[REWORD]` line and check it yourself.** This is the only place
   the rule can break, and it breaks politely.
3. **Check the friendly headings too.** Sections called things like *what sets
   you up for success here* are requirements wearing a nicer jumper.
4. **Check the filename and the document properties** before you attach it.
   `tools/cv-audit.py` does this for you. Another employer's name left in the
   properties is invisible on screen and fatal on arrival.
5. **Check every number against every other number.** Years, team sizes,
   percentages. The audit script prints them side by side so you can.

## The test that matters

For every claim on the page, predict the follow-up question. If the answer is a
story you can tell, the claim stays. If the answer is a careful sentence, the
claim comes out.

A CV that survives the second question is worth more than a CV that wins the
first one.