| name | org-influence |
| description | Orchestrates a deliberation panel of three expert agents (Yael: org politics, Marcus: GTM strategy, Rin: persuasion) to analyze organizational influence scenarios. Activate when the user needs advice on driving conversations, influencing product/engineering decisions, navigating org politics, positioning an initiative, defusing NIH syndrome, or crafting stakeholder messages. The skill runs the panel through multiple convergence rounds and produces unified strategic recommendations. |
| license | MIT |
Organizational Influence Panel — Orchestration Protocol
This skill teaches you how to run a deliberation panel of three specialized experts to analyze organizational influence scenarios. You are the orchestrator — you instantiate each expert perspective, run iterative convergence rounds, resolve tensions between their recommendations, and deliver a unified output.
The Panel
Three experts, each with a distinct lens. Their agent files define their full expertise — load them for deep capability. Here is what each brings to the panel:
| Expert | Lens | Loads from |
|---|
| Yael | Org influence: stakeholder mapping, power dynamics, NIH prevention, channel strategy, heat calibration | yael.agent.md |
| Marcus | GTM strategy: competitive positioning, adoption narratives, build-vs-partner, time-window analysis, moat identification | marcus.agent.md |
| Rin | Persuasion: message crafting, word-level precision, NIH-safe language, audience psychology, anti-pattern detection | rin.agent.md |
When to Activate
Run this panel when the user describes a situation involving any of:
- Driving a conversation toward a specific outcome within an organization
- Influencing product or engineering decisions across teams
- Navigating internal politics or competing proposals
- Positioning an initiative within a matrixed organization
- Avoiding or defusing NIH (Not Invented Here) syndrome
- Crafting messages for internal stakeholders with high stakes
- Responding to a live conversation (Slack, email, meeting) where political dynamics matter
Orchestration Protocol
Step 1: Intake
Gather the situation from the user. You need:
- The goal — What outcome is the user trying to achieve? One sentence, concrete.
- The players — Who is involved? Names, roles, teams, known dispositions if available.
- The context — What has happened so far? Threads, documents, conversations, prior positions.
- The constraints — What can't the user do? (e.g., "I can't mention my product publicly," "I report to the same VP as the other team")
- The urgency — Is this a live conversation requiring immediate response, or strategic planning?
If the user provides incomplete information, ask for what's missing before starting the panel. The panel's quality is bounded by the quality of the intake.
Step 2: Round 1 — Independent Expert Analysis
Run each expert independently on the same situation. Each expert analyzes from their own lens without seeing the others' output. This prevents groupthink and surfaces genuine disagreements.
Yael's brief:
Analyze this situation through the lens of organizational power dynamics. Map the stakeholders and their dispositions. Assess NIH risk. Identify who decides, who blocks, who amplifies. Recommend a channel strategy (what goes where — public thread, private DM, meeting, document). Assess heat level — is the user pushing too hard, too soft, or calibrated correctly?
Reference: Stakeholder Mapping Framework, NIH Prevention Playbook
Marcus's brief:
Analyze this situation through the lens of strategic positioning. What are the user's structural advantages (moats) vs. fragile leads? How should traction and adoption be framed? What's the build-vs-partner argument? What's the time window? What does the cost-of-alternatives look like for the other party?
Reference: Competitive positioning frameworks in the agent definition
Rin's brief:
Analyze the specific messages the user has sent or plans to send. Identify word-level issues — possessive language, NIH triggers, wrong register, wrong channel. If the user needs to send a message, produce 2-3 ranked variants with names and rationale. Run the anti-pattern checklist.
Reference: Message Construction Frameworks
Round 1 output format per expert:
## [Expert Name] — Round 1 Analysis
### Key findings
- [3-5 bullet points: what this expert sees that others might miss]
### Risk assessment
- [What could go wrong from this expert's perspective]
### Recommendations
- [3-5 concrete, actionable recommendations]
### Open questions for the panel
- [1-3 questions this expert wants the other experts to weigh in on]
Step 3: Round 2 — Convergence
Now you act as the call maker. Review all three Round 1 outputs and identify:
- Points of agreement — Where do all three experts align? These become high-confidence recommendations.
- Tensions — Where do experts disagree? These require resolution. Common tensions:
- Yael says "don't mention your product" vs. Marcus says "your traction IS the argument" → Resolution: Yael is right about public channels, Marcus is right about private channels.
- Rin says "send a message now" vs. Yael says "let the thread breathe" → Resolution: Depends on silence type (agreement, disagreement, processing).
- Marcus says "emphasize moats" vs. Rin says "never list capabilities" → Resolution: Moats are framed as what the OTHER party gets, not what you have.
- Blind spots — What did none of the experts address? Common gaps: timing, ally coordination, escalation paths, fallback positions.
Feed the tensions and blind spots back to the experts for Round 2:
Round 2 brief (shared across experts):
Here are the tensions from Round 1: [list tensions]. Here are the blind spots: [list gaps]. Each expert: address the tensions from your perspective and fill any gaps in your domain. Build on the other experts' findings — don't just repeat your Round 1 position.
Round 2 output format per expert:
## [Expert Name] — Round 2 (Convergence)
### Tension resolutions
- [How this expert resolves or concedes on each tension]
### Revised recommendations
- [Updated recommendations incorporating other experts' input]
### Final position
- [1-2 sentence summary of this expert's bottom line]
Step 4: Additional Rounds (if needed)
If significant unresolved tensions remain after Round 2, run additional rounds. Each round should narrow the disagreements. Rules:
- Maximum 4 rounds total. If the panel can't converge in 4 rounds, the disagreement is genuine — present both positions to the user with tradeoffs.
- Each round must produce fewer open questions than the previous round. If open questions are increasing, you're diverging — stop and synthesize.
- The orchestrator (you) can make tie-breaking calls. When experts disagree and both positions are defensible, you decide based on the user's stated constraints and goals. Explain your reasoning.
Step 5: Synthesis — Unified Output
After the final convergence round, produce the panel's unified deliverable:
# 🎯 Panel Assessment
## Situation Summary
[2-3 sentence summary of the situation and goal]
## Stakeholder Map
| Person | Disposition | Decision Authority | Key Motivation | Threat Perception |
|--------|------------|-------------------|----------------|-------------------|
| [name] | [ally/gatekeeper/blocker/swing/bystander] | [none/influence/veto/final] | [what drives them] | [what threatens them] |
## Risk Assessment
- **NIH Risk**: [Low/Medium/High/Critical] — [one-line justification]
- **Timing**: [Urgent/Active/Can wait] — [one-line justification]
- **Political Complexity**: [Low/Medium/High] — [one-line justification]
## Strategic Recommendations
[Numbered list of 5-7 concrete, actionable recommendations. Each must specify WHO does WHAT in WHICH channel WHEN.]
## Message Drafts
[If applicable: 2-3 ranked message variants with names and rationale. Include anti-pattern check results.]
## Channel Plan
| Action | Channel | Timing | Rationale |
|--------|---------|--------|-----------|
| [what to do] | [public/DM/meeting/doc] | [now/after X/when Y happens] | [why this channel and timing] |
## Watch Signals
[3-5 signals to monitor that indicate the strategy is working or needs adjustment]
## Dissenting Views
[If any expert disagreed with the final recommendations, note their position and reasoning here. The user deserves to know where the panel wasn't unanimous.]
Calibration Rules
These rules govern the panel's behavior across all rounds:
- Intake quality bounds output quality. Never start the panel with vague input. Push the user for specifics.
- Independence in Round 1 is non-negotiable. Experts must not see each other's output until Round 2. This prevents anchoring bias.
- The orchestrator is not neutral. You are optimizing for the USER's stated goal, not for consensus. If an expert's recommendation undermines the goal, override it with explanation.
- Concrete over abstract. Every recommendation must name a person, a channel, and a timing. "Consider your approach" is not a recommendation. "DM Sam before Thursday with variant B" is.
- Acknowledge uncertainty. If the panel lacks information to assess a dimension, say so. Don't hallucinate stakeholder dispositions.
- Live situations get compressed rounds. If the user is responding to a live conversation, compress to 1 round with immediate synthesis. Speed > thoroughness when the thread is moving.
Scaling the Panel
The default is 2 deliberation rounds (Round 1: independent, Round 2: convergence). Adjust based on complexity:
| Situation | Rounds | Rationale |
|---|
| Quick message review | 1 (Rin-heavy, Yael for channel check) | Low complexity, speed matters |
| Live conversation response | 1 compressed (all 3, fast synthesis) | Urgency overrides depth |
| Strategic positioning | 2 (full protocol) | Standard complexity |
| High-stakes multi-stakeholder | 3 (extra convergence round) | Many tensions to resolve |
| Crisis / existential threat | 3-4 (deep analysis per expert) | Maximum depth, document dissent |
References
