菜单
The single outbound chokepoint for every autonomous action Bravo performs on behalf of CC. Enforces CASL compliance, cooldown windows, daily caps, and cross-engine idempotency architecturally — callers cannot bypass it.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Higgsfield image prompt director covering Banana Pro, Soul Cinema, and GPT-2. Three asset types in strict order — single-image character outfit on white seamless studio (Banana Pro or Soul Cinema two-step), 6-panel multi-angle character sheet off that base, and scene plates with/without characters. Plus GPT-2 for detail face/chest-up portraits. Reads reference images for hair, makeup, wardrobe, jewelry, identity. Outputs photorealistic prompts with locked hyperreal stack — skin pores, subsurface scattering, strand-by-strand hair, fabric weave, Kodak Vision3 film. Use for any character outfit reference, character sheet, model sheet, ref sheet, multi-angle sheet, scene plate, environment plate, face detail, or photorealistic still — even without saying 'Banana Pro,' 'Nano Banana,' or 'Soul Cinema.'
Universal cinema worldbuilding director for Seedance video prompts. Five cinema modes (Narrative, Studio, Action, Performance, Atmospheric) with locked ARRI + Panavision/Cooke camera specs and diegetic-only audio rules. Reads reference images for wardrobe, hair, makeup, identity, and environment. Outputs production-ready Seedance prompts. Use whenever CC asks for a Seedance video prompt, mentions Seedance, asks for cinematic scene breakdowns, uploads reference images for a scene, describes a scene for video generation, or asks for shot prompts for music videos, action sequences, performance scenes, narrative shorts, fashion films, or atmospheric environment plates — even if 'cinematic' or a mode name isn't said.
Complete automated video production system for CC's content pipeline. Use whenever CC provides raw video footage and says "make this a post", "edit this", or asks for a cinematic output. Replaces a human video editor entirely. Covers hook engineering, pacing, caption styling, color grading, audio mastering, sound design, motion graphics, and multi-platform export.
Read-only social + web listening across public platforms (YouTube, Twitter/X, LinkedIn, Reddit, the open web, GitHub, Exa search) for competitive intelligence, content research, transcript mining, and checking public post results. Wraps Agent Reach via scripts/social_reach_tool.py. Does NOT read Instagram, TikTok, or Facebook.
Build high-converting, education-first advertorial / "infomercial" landing pages for ecommerce products — long-form scrollable HTML that warms cold paid traffic and clicks through to the product page. Covers angle selection, direct-response copy doctrine, AI image generation with a product reference, a premium anti-slop design system, and a portable build/QA convention. Use for advertorials, pre-sell pages, native/listicle ad landers, VSL pages, or any "read-an-article-then-buy" funnel page.
Automated video editing pipeline for CC's personal brand content. Use whenever CC drops raw footage and needs it edited, captioned, graded, and sent for review. The primary tool is video_editor.py which handles the full 8-step pipeline.
| name | send-gateway |
| description | The single outbound chokepoint for every autonomous action Bravo performs on behalf of CC. Enforces CASL compliance, cooldown windows, daily caps, and cross-engine idempotency architecturally — callers cannot bypass it. |
| disable-model-invocation | true |
| triggers | ["route outbound through gateway","send via send_gateway","enforce cooldown before sending","check if lead was recently contacted","outbound chokepoint"] |
Rule of law: If code sends email, a DM, or any outbound message on CC's behalf, it goes through
scripts/send_gateway.py. Directsmtplib.SMTP_SSL()calls from a business engine are a regression and must be reverted in review.
Before 2026-04-20, four Python engines and one N8N workflow could contact the same lead on the same day without seeing each other. They wrote to three different tables. No engine could answer "was this lead contacted in the last 72 hours?" before acting. That is the root cause of the duplicate-email bug CC described in the 2026-04-19 audit.
This skill encodes the rule that closes that bug: every send goes through send_gateway.send(). Idempotency is not a library callers must remember to invoke — it is architecturally enforced because the smtplib call lives inside the gateway and nowhere else.
Use the gateway whenever your code:
Do NOT bypass the gateway just because a send is "simple" or "one-time." The whole point is that the cooldown ledger remains whole.
from send_gateway import send
result = send(
channel="email", # email | instagram | linkedin | phone | skool
agent_source="outreach_engine", # identify the caller (free text, stay on known values)
to_email="jane@acme.example", # for email channel
lead_id=None, # auto-resolved or auto-created from to_email
subject="Quick question about your HVAC scheduling",
body_text="Hi Jane, ...",
body_html=None, # optional
brand="oasis", # oasis | conaugh_mckenna | nostalgic
intent="commercial", # commercial | transactional | internal
cooldown_hours=None, # None = DEFAULT_COOLDOWNS[channel]
metadata={"campaign": "hvac-q2-2026"}, # free-form dict persisted in ledger
ics_content=None, # .ics string for calendar invites
ics_filename="meeting.ics",
dry_run=False,
)
# result is always a dict with the same keys:
# {"status": "sent"|"blocked"|"suppressed"|"dry_run"|"error",
# "reason": str, "lead_id": str|None, "interaction_id": str|None,
# "cooldown_until": str|None, "daily_count": int|None}
send() NEVER raises. On any error it returns status="error" with a reason string. Callers can rely on the return shape.
python scripts/send_gateway.py --json send --channel email \
--to jane@acme.example --subject "..." --body "..." \
--agent-source manual_cc
python scripts/send_gateway.py can-act --lead-id <uuid> --channel email --json
python scripts/send_gateway.py history --lead-id <uuid> --limit 10
python scripts/send_gateway.py stats --json
| Intent | Suppression check | CASL footer | List-Unsubscribe | Example use |
|---|---|---|---|---|
commercial | enforced (fail-closed) | added | added | cold outreach, nurture, sales |
transactional | skipped (CASL s.10(9)) | added | added | booking confirmations, reminders, password resets |
internal | skipped | skipped | skipped | internal test mail, CC-to-self notifications |
The brand keyword selects CASL footer sender name + business name + address. Add a brand to BRAND_IDENTITY in scripts/send_gateway.py when CC wants a new one to share the chokepoint.
| Brand | Business name | Sender | Use for |
|---|---|---|---|
oasis (default) | OASIS AI Solutions | Conaugh McKenna | Agency outreach, client comms |
conaugh_mckenna | Conaugh McKenna | CC (Conaugh McKenna) | Personal brand, content, DJ inquiries |
nostalgic | Nostalgic Requests | Conaugh McKenna | Nostalgic Requests product mail |
Conservative by default — CC is still building reputation; better to under-send than look spammy.
| Channel | Cooldown | Daily cap |
|---|---|---|
| 72h (3 days) | 50/day | |
| 48h | 30/day | |
| 72h | 20/day | |
| phone | 168h (7 days) | 15/day |
| skool | 24h | — |
| telegram | 0 (internal) | — |
Override per-call with cooldown_hours=<int>. Daily caps are hard — gateway refuses over-cap sends with status="blocked".
All five outbound Python engines now route through the gateway:
| Engine | Function that calls send() |
|---|---|
| outreach_engine.py:send_outreach() | cold outreach with Meet invite .ics |
| outreach_batch.py:send_approved_draft() | Telegram-approved batch sends |
| email_engine.py:cmd_send() / cmd_send_template() / cmd_sequence_run() | one-off, templated, and sequence sends |
| funnel_nurture.py:send_email() | Day 2 / Day 5 follow-ups |
| booking_engine.py:_send_booking_confirmation() / _send_reminder_email() | transactional confirmations + reminders |
Every successful send writes to three places:
lead_interactions (architectural truth — cooldown_until + agent_source + metadata)email_log (legacy SMTP-layer truth — keeps analytics and report tooling working)leads.last_contacted_at (keeps CRM pipeline view fresh)Failed sends write only to email_log with status='failed' for forensics.
cooldown_until, agent_source, metadata columns + indexes. Apply via python scripts/apply_migration.py.get_entity_context(lead_id) returns the full relationship context for persona-aware LLM drafting.KNOWN_CHANNELS (frozenset).DEFAULT_COOLDOWNS.DAILY_CAPS.send(), add a channel-specific branch that performs the physical send — route it through a _send_<channel>() helper that is the ONLY place where the real client call lives.Maven owns scripts/send_gateway.py for marketing email blasts, Meta Ads spend, Google Ads spend, and any DM/social-post outbound. The architecture is identical to Bravo's — single chokepoint, fail-closed critic gate, name sanitization, daily/hourly caps, suppression-list check. Caps are stricter on the Maven side because marketing list sizes are larger (200/day vs Bravo's cold-outreach pace). The killswitch env var is MAVEN_FORCE_DRY_RUN=1. Paid-spend channels (meta_ads, google_ads) additionally consult cfo_pulse.json before approving. Maven NEVER bypasses the gateway, even for "small" tests — every outbound action goes through it or it's a regression.