| name | ddd-pollinate |
| description | Express THIS DDD's product value to audiences — message-first, format follows — via a bundled, decoupled copy of SwarmAI's real pollinate engine (11 format tracks + scripts + moat gates), sourcing value from the DDD's own ② PRODUCT.md, file-based, no SwarmAI backend. DDD-native decouple of s_pollinate.
TRIGGER: "ddd pollinate", "express this ddd's value", "make content from this ddd".
NOT FOR: SwarmAI's own s_pollinate (bound to Swarm brand + SwarmWS paths). |
| tier | lazy |
DDD-Native Pollinate (s_ddd-pollinate)
The value-expression capability a DDD carries: turn what the product IS (its ②
Knowledge — vision, differentiation, proof) into audience-shaped media.
Message-first, format follows the audience.
DDD-native decouple of SwarmAI's s_pollinate. This is NOT a hand-written shell —
it bundles the REAL engine (11 format tracks, ~25 scripts, the moat gates) copied from
s_pollinate and decoupled: every hardcoded ~/.swarm-ai path resolves from
$SWARM_WORKSPACE, SwarmAI's brand assets are stripped (you supply your own via
brand/identity.yaml), and the publish target + brand checks are configurable — so a
DDD runs it on Kiro / Claude Code / an AIM package with no SwarmAI backend.
What's bundled (real engine, not a stub)
| Part | What | Portable? |
|---|
tracks/ (11) | track-a-video … track-k-podcast — the format playbooks | ✅ verbatim |
references/, REVIEW_PATTERNS.md, phonemes.json | platform specs, review rubric, TTS phonemes | ✅ verbatim |
brand/identity.yaml | TEMPLATE — fill with YOUR project's name/colors/voice | ⚙️ you fill |
data/slide_bold_previews/ | 34 deck design systems (bundled, was in s_frontend-design) | ✅ bundled |
scripts/ | evaluate/format-recommend/validator/gates/tts/shorts/publish | ✅ decoupled |
| the moat gates | pollinate_validator.py + convergence_gate.py + cross_format_check.py + p2_scan.py | ✅ pure stdlib |
The principle (retained)
Do NOT start from "make a poster." Start from: who is the audience, what is the ONE
message that changes their mind, THEN pick the format that lands it. Source every
claim from THIS DDD's ② PRODUCT.md — never invent proof not in ② Knowledge.
THE MOAT (option C — validator-run, agent-enforced)
Unlike the pipeline, pollinate has no artifact_cli chokepoint — its moat is run the
validator as a BLOCKING step (the same way the poster track already runs p2_scan):
python scripts/pollinate_validator.py <content_dir> --json
pollinate_validator.py (9 structural invariants incl. cross-format 7-9) +
convergence_gate.py (8-layer poster gate) + p2_scan.py (anti-slop first-person scan)
are ALL pure stdlib and run standalone. The agent MUST run the validator before DELIVER
and fix every error — degrading to "review this adversarially" only where no runtime can
execute it. Never silently skip.
Brand checks are DDD-configurable, NOT SwarmAI-locked (opt-in via env):
POLLINATE_REQUIRE_QR=1, POLLINATE_REQUIRE_LINK=<substr>, POLLINATE_WATERMARK=<text>.
Unset → those checks auto-pass (a portable DDD isn't forced to carry SwarmAI's QR/watermark).
Deliver boundary
Emit a portable package under <workspace>/.artifacts/pollinate/output/<slug>/; the
human/runtime publishes it. Publishing needs POLLINATE_PUBLISH_REPO=<owner>/<repo> (YOUR
repo — never a hardcoded one). NEVER auto-post to an external channel.
See INSTRUCTIONS.md for the full stage flow.
