career-synthesize

Files (1)

• SKILL.md

SKILL.md (first lines)

# /career-synthesize - turn evidence into a position

Rebuilds `PROFILE.md`. Run it after any meaningful ingest or interview, and any time the user asks what they are actually good at.

Flags: `--shareable` produces `PROFILE-shareable.md` with third-party quotes anonymised and private material removed.

## Read everything first

All of `Roles/`, `Wins/`, `Praise/`, `Skills/`, `People/`, plus `Preferences.md` and `STATE.md`. This command is worthless done partially, because its entire value is seeing patterns across items that were written in isolation.

If the evidence is thin, still produce a profile. Mark it `confidence: draft`, make the holes visible, and say what would most improve it. A rough profile with obvious gaps motivates people to fill them; a request for more input does not.

## Finding the position

Positioning is not a summary of everything they have done. It is the small number of things that are both **rare** and **valued**, stated so that the right reader recognises themselves as the buyer.

Work through it in this order:

1. **Theme the praise.** Group `Praise/` by what it is evidence of, ignoring who said it. Themes that recur across different people, different companies, and different years are the strongest signal in the entire vault. One person saying it is an opinion. Five people across three employers saying the same thing over six years is a fact about them.

2. **Weight unsolicited over solicited.** A LinkedIn recommendation is written to be flattering. An offhand line in a Tuesday standup is not. Weight accordingly, and prefer the offhand line when quoting.

3. **Find the counterfactual skills.** Go through `Wins/` asking which would have happened anyway with a competent replacement, and which needed this specific person. The second set is the position. Most people lead with the first set, because it is bigger.

4. **Name the perception gap.** Compare what the praise says they are good at against what they say they are good at. The gap is almost always the most useful output of this whole exercise. People systematically undervalue what comes easily to them, and the thing that comes easily is frequently the thing others cannot do at all. State the gap explicitly and in plain language.

5. **Find the combination, not the skill.** Very few people are the best in the world at one thing. Plenty are unusual at a combination: ships production code and can hold a keynote; runs a community and reads a P&L; teaches beginners and debugs kernels. The combination is where the defensible position lives, and it is what makes a role search specific instead of a scattergun.

6. **Check it against the preferences.** A position built on work they hate is a trap with a salary attached. If the strongest evidence points somewhere `Preferences.md` says they do not want to go, do not quietly optimise for the evidence. Surface the tension and let them decide.

## Writing PROFILE.md

Sections, in this order:

**In one line.** What they do, for whom, and why it matters. It should be specific enough that it excludes people. If it could describe half the field, it has failed.

**What you are demonstrably good at.** Three to five claims, no more. Each with the evidence beneath it: the wins that show it and the quotes that corroborate it, each linked. Each tagged `strong`, `moderate`, or `thin`. Ruthless selection is the point - a list of twelve strengths communicates nothing, and the discipline of cutting to four is where the value is created.

**What people say about you.** Quotes, grouped by theme, attributed with name, relationship, and date. Lead with unsolicited. This section is also what they read on a bad day, and that is a legitimate function of it.

**How you work.** Conditions where they do their best work and where they do not, drawn from `Preferences.md` and corroborated by patterns in `Roles/`. Written as observations, not aspirations.

**Role archetypes to target.** Three to five, ranked. For each: what the role is, why the evidence supports it, where the stretch is, what would make them a strong candidate, and what is currently missing. Be concrete about the stretch. In developer relations the usual candidates are developer advocate, developer experience engineer, developer educator, community lead, developer relations manager or head of, solutions engineering, technical product marketing, principal or field roles, and a return to engineering. Do not list all of them. Pick the ones the evidence actually supports and say why the others were excluded, because the exclusions are as useful as the picks.

**What is still thin.** Claims that currently rest on one source, no numbers, or self-report only. For each, name the specific thing that would firm it up. Mirror this list into the `thin claims` section of `STATE.md`.

## Preserve their edits

If `PROFILE.md` already exists and has been hand-edited, keep their wording. Diff against what you would have written, update what is factually stale, and leave phrasing alone. Where new evidence changes a conclusion they wrote themselves, do not silently overwrite it - flag it and ask.

## Then keep going

Turn every thin claim and every open contradiction into a ranked entry in the `STATE.md` open questions queue, with the reason it matters.

Then respect the intensity dial. At `normal` or `high`, offer to work through the top two or three gaps immediately, since motivation is highest right after seeing the profile. At `low` or `paused`, present the profile, note that the gaps are logged, and leave it.

Update `last_synthesized` and the coverage table. Commit.

## Reporting

Do not paste the whole profile into chat. Give them, in prose: the one-line position, the perception gap, the top role archetype and why, and the single most valuable gap to close. Then point them at the file.

If this run materially changed the position from the previous one, lead with that and say what caused the change.

From bendechrai/career-brain, MIT.

More from bendechrai/career-brain