branding-assessment

name	branding-assessment
description	Audit the user's public brand surfaces — personal site, GitHub, Hugging Face, LinkedIn — against ground-truth direction. Surfaces drift (site says X, ground-truth says Y) and gaps (no project shipped in domain Z). Produces an actionable update list.
disable-model-invocation	false
allowed-tools	Read, Write, Bash(mkdir *), Bash(test *), Bash(date *)

Branding Assessment

Surface mismatch between what the user says they're focused on (ground-truth) and what their public presence shows. Hiring managers and prospects look at the surfaces; ground-truth is invisible.

This is a diagnostic skill — it produces a list of edits, doesn't apply them.

Inputs

$ARGUMENTS:

• Optional: --surface=<website|github|hugging-face|linkedin|all> — default all.
• Optional: --target=<role|domain|consulting> — what the brand should be optimised for. Defaults to ground-truth's primary direction.
• Optional: --depth=<quick|deep> — default quick (top-of-page only). deep walks into individual repos / projects.

Prerequisites

• ground-truth.md populated.
• branding/config.yaml (or workspace config keys):
  • WEBSITE_URL
  • GITHUB_HANDLE
  • HUGGING_FACE_HANDLE
  • LINKEDIN_URL

If config missing, prompt for the URLs and offer to write branding/config.yaml.

Procedure

1. Load context

• ground-truth.md (direction, domains-of-interest, freelance-brand, hard-constraints).
• branding config.

2. Fetch each surface

For each surface in scope:

• Website — fetch home page + about page. Extract: tagline, primary positioning, recent posts/projects, calls to action.
• GitHub — profile bio, pinned repos, recent commit cadence, language distribution. --depth=deep reads the README of pinned repos.
• Hugging Face — profile bio, models/datasets/spaces, recent activity.
• LinkedIn — headline, about, recent activity (only what's publicly accessible without auth).

If a surface fetch fails, note it; don't bail.

3. Compare to ground-truth

Per surface, score on three axes:

• Direction match — does the surface's positioning match ground-truth's primary direction?
  • aligned / partial / drifted / silent.
• Domain coverage — are ground-truth's domains-of-interest visible?
  • well-covered / partial / missing.
• Freshness — last update vs. cadence target.
  • fresh / stale (60–180 days) / very stale (>180 days).

4. Surface findings

Categorise:

• Critical — direction drifted (e.g. ground-truth = "AI policy consulting", website tagline = "full-stack dev for hire").
• Gaps — domain in ground-truth but no shipped artifact in that domain on any surface.
• Stale — surface hasn't moved in months while user has shipped elsewhere.
• Quick wins — bio one-liners, pinned-repo selection, headline edits.

5. Write report

${WORKING_FOLDER}/branding/assessment-<date>.md:

# Branding assessment — {{date}}

**Target:** {{role|domain|consulting}}
**Surfaces:** {{list}}

## Direction match

| surface | score | evidence |
| --- | --- | --- |
| website | partial | tagline says X, ground-truth says Y |
| ...

## Domain coverage

(per domain → which surfaces show shipped work)

## Stale surfaces

...

## Recommended edits (ordered)

1. **{{surface}}** — {{specific edit}}. Why: {{from-ground-truth}}.
2. ...

## Gaps to ship

(domains in ground-truth with no public artifact — candidate `/career:suggest projects` queries)

6. Print summary

report: branding/assessment-<date>.md
critical: <n>  gaps: <n>  stale: <n>  quick-wins: <n>

next:
  critical → edit the offending surface
  gaps → /career:suggest projects --domain=<missing>

Guardrails

• Recommendations only; never edits external surfaces.
• Don't moralise. "Stale GitHub" is a fact, not a failure.
• Specific evidence required for every finding (a quoted line, a missing-pinned-repo note). No generic "your brand is unclear."

Failure modes

• Surface unreachable (auth wall, 404) → note and continue.
• ground-truth absent → bail; assessment is meaningless without an anchor.