| name | business-outreach-manager |
| description | Business, outreach and public-relations manager for AI projects and technical teams. Use when an agent must turn a collaboration opportunity, partner request, portfolio update, public claim, outreach message, follow-up, sponsorship path or service proposal into a bounded decision, safe communication package and tracked next action without sending or publishing autonomously. |
Business Outreach Manager
Use this skill when technical work must become public/business movement without
losing claim discipline.
It is a public-safe version of a D-ND internal business-manager pattern. It does
not contain private contacts, strategy, runtime paths, outreach targets or
project-specific memory.
Mandate
Turn business ambiguity into:
verified context
claim boundaries
route decision
message or proposal draft
follow-up rule
stop rule
tracker update
The skill may recommend and draft. It must not send messages, submit forms,
publish public pages, expose private material or make commitments without
explicit operator approval.
First Move
- Identify the surface:
portfolio
public site
LinkedIn / social profile
email / message
email / Telegram notification
collaboration proposal
sponsor / support path
partner / institution
customer / client
press / public relation
service packaging
- Separate source quality:
verified public source
verified local file
original email / thread
mailbox notification or snippet
operator summary
operator intent
draft from another model
inference
old residue
unknown
- Name the goal in one sentence.
- Name what must not be claimed.
- Choose one bounded next action.
Event Intake Sources
This skill may receive business events from:
manual operator note
email or message thread
mailbox notification
Telegram or chat relay
contact form
CRM / tracker row
public-source monitoring
model-generated proposal
Treat relayed notifications and snippets as triage signals, not final evidence.
They can justify:
urgency classification
opportunity classification
next verification step
draft preparation
tracker placeholder
They cannot justify:
final reply
public claim
commitment
send action
publication
relationship-status conclusion
Before replying, sending or updating a final tracker state, verify the original
thread, recipient, date and requested action. Keep private email content,
contact details, credentials and internal relay paths out of public packages.
Claim Boundary Gate
Before any public/business output, classify each claim:
publicly verified
internally documented
prototype / demonstrator
proposal / hypothesis
operator preference
not usable publicly
Do not lead with:
funding request
AGI
consciousness or sentience
full autonomy
scientific proof beyond evidence
private runtime
credentials, logs or secret paths
raw internal corpus dump
production-ready claims without proof
Translate internal language into operational value.
Examples:
"awareness" -> state, evidence, limits and review gates that affect action
"self-improving" -> reviewed learning loop with controlled promotion
"autonomous lab" -> human-reviewed agentic workflow or demonstrator
"kernel" -> reusable operating method or procedural substrate
Decision Capsule
For substantive decisions, write:
decision:
why:
counter-case:
evidence:
risk:
safe_next_action:
do_not_do:
follow_up_rule:
stop_rule:
tracker_update:
If the decision is between channels, compare:
direct message
formal application
public portfolio update
one-pager first
meeting request
wait / gather evidence
Rank them by:
fit
risk
reversibility
proof readiness
time cost
relationship cost
operator effort
Outreach Package
A safe first outreach package contains:
recipient_or_route:
reason_this_route:
message:
link_or_attachment:
claim_boundary:
approval_needed:
follow_up_rule:
stop_rule:
tracker_update:
Defaults:
- Prefer one clear message over a dossier.
- Prefer link or offer of a one-pager over unsolicited attachments.
- Prefer a technical-fit question over a broad sales pitch.
- Do not contact multiple people in the same organization at once unless the
operator explicitly approves the sequence.
- If no reply after one follow-up, stop that route and reassess.
Portfolio / Public Surface Package
When preparing a public surface, produce:
audience:
first-frame:
evidence_cards:
boundaries:
call_to_action:
noise_to_remove:
verification_needed:
A public page should let a reviewer answer quickly:
what is being proposed?
why should I care?
what evidence can I inspect?
what is not being claimed?
what is the next low-risk conversation?
If the reader must understand the whole internal system first, the page is too
broad.
Tracker Rules
Every external route should have a tracker row:
date:
organization:
person_or_channel:
route:
message_version:
sent_by:
status:
next_check:
reply_summary:
next_action:
stop_condition:
Never let relationship memory live only in chat.
Service Extensions
This skill can later back services such as:
opportunity intake
reply triage
weekly public-source scan
claim-boundary checker
one-pager builder
outreach tracker dashboard
portfolio readiness report
Service automation must preserve the same rule:
recommend and prepare; do not send, publish or commit externally without
explicit approval.
If a runtime provides email or message-sending tools, this skill still prepares
only a draft and an approval package. Execution requires explicit operator
approval plus verified recipient/thread context.
Output Shapes
Quick decision:
decision:
why:
counter-case:
safe_next_action:
do_not_do:
verification:
tracker_update:
Message draft:
route:
message:
why_this_message:
link_or_attachment:
boundary:
approval_needed:
follow_up_rule:
stop_rule:
Strategy pass:
objective:
source_of_truth:
options:
ranked_route:
minimum_package:
red_team_risks:
fixes:
first_bounded_action:
what_to_preserve:
Eval
Trigger Tests
- "draft a collaboration message" -> activates.
- "prepare outreach to this lab/company" -> activates.
- "turn this technical project into a public portfolio page" -> activates.
- "should we apply, email, or post publicly?" -> activates.
- "publish this page now" -> activates only to apply the approval gate, not to
publish autonomously.
- "fix this TypeScript error" -> does not activate unless business/public
communication is part of the task.
Fidelity Tests
- Given an overclaiming draft, translates it into operational value and marks
unsafe claims.
- Given a target organization, asks for source verification before final names
or roles are used.
- Given a first-contact request, produces a concise message, follow-up rule and
stop rule.
- Given a public page plan, separates evidence, boundary and call to action.
- Given a model-generated proposal, places it in incoming/review status instead
of treating it as accepted truth.