brief-cultural-fit

name	brief-cultural-fit
description	Produce a Cultural Fit research brief — pairs the user's stated operating style and constraints (read from ground-truth.md) against the company's apparent operating mode (from public signals). Output is a match-map plus blunt friction list. Use after the other briefs are in hand. Requires populated ground-truth.md. Reads templates/research-briefs/cultural-fit.md and writes to <WORKING_FOLDER>/research/companies/<slug>/cultural-fit.md.
disable-model-invocation	false
allowed-tools	Read, Write, WebSearch, WebFetch, Bash(mkdir *), Bash(test *)

Brief — Cultural Fit

The brief that pairs user's operating style with the company's. Most consequential for retention, least consequential for getting an interview. Run it last — it benefits from the other briefs already being on disk.

Inputs

$ARGUMENTS:

• Required: <company_name>.
• Optional: --slug=<slug>.
• Optional: --lens=<employer|client|partner> (default employer).

Procedure

1. Resolve config + paths

${CAREER_DATA_DIR}/config.json → WORKING_FOLDER. Output: ${WORKING_FOLDER}/research/companies/${slug}/cultural-fit.md.

2. Read inputs

• Template: ${PLUGIN_ROOT}/templates/research-briefs/cultural-fit.md.
• Required: ${WORKING_FOLDER}/ground-truth.md. If missing or essentially empty, prompt the user to run /career:ground-truth first; do not proceed with placeholder values.
• If they exist, also read other briefs in the same dir (company-overview.md, glassdoor-signal.md, recruitment-profile.md) for input.

3. Pull user's stated style

From ground-truth.md:

• Looking for (employment / contract / both).
• Hard constraints (location, hours, family, other).
• Domains of interest.

If ground-truth doesn't cover collaboration preferences / decision-making style / what drains vs energises: ask the user inline. Three short questions. Don't fabricate.

4. Read company's operating mode

From public signals — engineering blog posts, podcast appearances, public Slack/Discord activity, careers-page values, Glassdoor culture themes, public memos. Cite each.

Categorise:

• Communication mode (written-first / meeting-first / sync-heavy / async-first).
• Decision velocity.
• Hierarchy.
• Operating cadence.
• On-call expectations.
• Documented values + whether they appear operationalised.

If glassdoor-signal.md exists in the workspace, read its theme analysis as a primary input. Don't re-research what's already on disk.

5. Match map

Per-dimension table comparing user vs company. Fit values: match | tension | mismatch | unknown.

6. Outright frictions

Hard mismatches with concrete examples. Be blunt.

7. Soft frictions / things to probe

Open questions that would resolve uncertain verdicts. Frame as questions the user would ask in an interview / discovery call.

8. Verdict

• Cultural fit: strong | likely | uncertain | poor
• Confidence: high | medium | low
• Headline reason: one sentence.

Decision rules:

• strong requires at least one specific evidence pair (user dimension → company signal).
• poor is valuable; do not soften.
• uncertain is acceptable when ground-truth is sparse or company signals are weak.

9. Write output + update CRM

Append +cultural to the row.

Companion delegation

If decision-evaluation-framework is installed, the brief output is consumable by decision-evaluation-framework:swot if the user wants a structured pivot/decline analysis later.

Guardrails

• Ground-truth is privileged — don't override with company marketing.
• "Strong fit" requires evidence pairs. Generic praise is not allowed.
• "Poor fit" is not softened.
• Run last among the six briefs; benefits from the other outputs.

Failure modes

• ground-truth.md missing or empty → bail; prompt to run /career:ground-truth.
• No public signals available (stealth company) → write ## Status: low-signal — need to ask directly; produce the questions section without a verdict.

Idempotency

• Re-run prompts before overwrite.
• --refresh silently overwrites.