| name | act-brand-alignment |
| description | ACT brand alignment for all ecosystem projects. Use for ANY work on ACT sites, projects, content, design, or strategy. Understands ACT identity, LCAA method, dual-entity structure, all projects, voice/tone, and visual language. |
ACT Brand Alignment
When to Use
- Writing/editing ANY ACT content (web, marketing, grants, reports)
- Designing UI, visual assets, or brand materials
- Planning information architecture
- Creating content models or data structures
- Reviewing copy for voice/tone alignment
Quick Reference
Identity
A Curious Tractor is a regenerative innovation ecosystem partnering with marginalised—especially First Nations—communities to dismantle extractive systems. Like a tractor's power take-off, we transfer resources to community-led initiatives. We hand over the keys.
Method (LCAA)
Listen → Curiosity → Action → Art
Values
- Radical Humility (no saviors)
- Decentralised Power (communities lead)
- Creativity as Disruption (revolution through imagination)
- Truth-telling (name extractive systems)
Promise
40% of profits flow to community ownership. We design for our own obsolescence.
Project Scope Mapping
| Project | Focus | Load |
|---|
| Hub (act.place) | ACT as ecosystem, LCAA, partnerships | brand-core.md |
| ACT Farm / BCV | Land practice, conservation, residencies | land-practice.md |
| JusticeHub | Justice innovation, forkable models | projects-ecosystem.md |
| Empathy Ledger | Ethical storytelling, consent, sovereignty | projects-ecosystem.md |
| The Harvest | CSA, community gatherings | land-practice.md |
| Goods on Country | Circular economy, waste-to-product, comms agent system | goods.md |
| Art | Art as revolution, cultural production | projects-ecosystem.md |
Voice Guardrails
Writing voice (load references/writing-voice.md for ANY public-facing writing). ACT voice uses Ian Curtis's method of compression: name the room, name the body, load the abstract noun, stop the line before the explanation. Forbidden AI tells (delve, crucial, pivotal, tapestry, underscore, em dashes, negative parallelisms, superficial -ing tails, "challenges and future prospects" formulas) are listed in that file and must be rejected on sight.
Auto-grade before sending
For any pitch, grant, web copy, board report, donor letter, journal spread, caption, or essay, run the voice grader before declaring the draft ready:
node scripts/grade-voice.mjs --file <path> --project <slug> --genre <slug>
--project slugs: hub, justicehub, empathy-ledger, goods, bcv, harvest, farm, art, oonchiumpa, bg-fit, mounty-yarns, picc
--genre slugs: pitch, grant, web, board-report, donor-letter, journal-spread, caption, essay, press
Verdict semantics:
pass → safe to send (all four Curtis moves landed, no AI tells, plainness clean)
warn → operational sections may carry technical register (acceptable for working drafts; tighten the cover/exec-summary/email opener before submission)
fail → at least one Tier 1 hard rule tripped (em-dash, forbidden vocab, negative parallelism, etc.) — do not send until fixed
The grader writes its full output to stdout and exits non-zero on fail. Capture grades to thoughts/shared/reviews/<slug>.voice-grade.md for any artefact going to a funder.
Rubric source of truth: thoughts/shared/rubrics/act-voice-curtis.md v1.0 (calibrated 6/6 on canonical fixtures). Any rule change must mirror back to references/writing-voice.md.
Cost: ~$0.02 per grade (Sonnet 4.6, ~3K tokens). Use --tier1-only for a free deterministic check.
Funder-cadence grade (in addition to voice grade, for any funder-facing draft)
For any pitch, term-sheet response, renewal, follow-up, or report addressed to a specific funder, also run the funder-cadence grader. Layered on top of the voice grade: voice-curtis catches AI-tells, funder-cadence catches funder-specific cadence misses (wrong opener for the relationship stage, leading with a claims_to_avoid claim, dollars cited without an invoice or signed-letter row, primary-contact name out of date).
node scripts/grade-funder-cadence.mjs --file <path> --funder <slug> [--cycle <slug>]
--funder slugs (resolved against wiki/narrative/funders.json): minderoo, qbe-catalysing-impact, dusseldorp-forum, paul-ramsay-foundation, tim-fairfax, smith-family, amnesty-australia, niaa, jcf, atlassian-foundation, snow-foundation, patagonia, allbirds, who-gives-a-crap, centrecorp, rotary-eclub-outback, vincent-fairfax, social-impact-hub, state-qld-dfsdscs, streetsmart-australia, westpac-scholars-trust
--cycle slugs (optional): intro, pitch, renewal, report, followup, term-sheet. Skip if the cycle isn't yet defined in funders.json.
--tier1-only for a free deterministic pass (skips the Sonnet 4.6 call but keeps every dollar-citation, forbidden-claim, and canonical-name check).
Verdict semantics mirror voice-curtis (pass / warn / fail); a Tier-3 judgment miss (claims-to-avoid leading, dollar uncited, stage misalignment, primary-contact stale) counts as fail not warn per the 2026-05-07 calibration. Capture grades to thoughts/shared/reviews/<slug>.funder-cadence-grade.md for any artefact going to a funder.
Run both the voice-curtis grader AND the funder-cadence grader before declaring a funder-facing draft ready to send. Voice grader is funder-agnostic (project + genre); funder-cadence is funder-specific (claims, stage, contact).
Rubric source of truth: thoughts/shared/rubrics/funder-cadence.md v0.1 (calibrated 6/6 on Minderoo Goods pitch, QBE term-sheet, Dusseldorp renewal + 3 negative fixtures).
Alignment-loop self-grade (for synthesis docs only)
Auto-applied by the three scripts/synthesize-*.mjs Phase-1 cycle scripts (project-truth-state, funder-alignment, entity-migration-truth-state). Grades the synthesis against thoughts/shared/rubrics/alignment-loop-synthesis.md v0.1 via scripts/lib/alignment-loop-grade.mjs. On pass the synthesis commits clean; on warn/fail a triage report lands at wiki/output/lint-loop-YYYY-MM-DD.md. Pass --no-grade only when running in environments without ANTHROPIC_API_KEY (cron stubs, CI without secrets).
DO:
- Farm metaphor: seeds, harvest, cultivating, soil, seasons, fields
- Community ownership, co-stewardship, Indigenous sovereignty
- "Designing for obsolescence" / "handing over the keys"
- Name Jinibara Country when referencing the land
- 40% profit-sharing commitment
- Concrete rooms. Concrete bodies. Stopped lines.
DON'T:
- Overclaim ("world-leading," "revolutionary")
- Extractive, luxury, or commercial language
- Frame communities as beneficiaries (they are co-owners)
- Corporate jargon or glossy marketing speak
- Imply permanence (we design for obsolescence)
- AI register (puffery, rule of three, elegant variation, em dashes, significance claims)
Visual Language
- Palette: Stone 50-900 + Emerald accents
- Typography: Geist Sans
- Motifs: Seeds, cockatoos, regenerative farming
File References
| Need | Reference |
|---|
| Identity, LCAA, values, voice | references/brand-core.md |
| Writing voice (Curtis method + AI tells to avoid) | references/writing-voice.md |
| Land details, conservation | references/land-practice.md |
| All projects/seeds, revenue | references/projects-ecosystem.md |
| Page patterns, IA | references/content-structure.md |