Comms-Stack

An AI-native communications operating system for Stripe Ireland.

Comms-Stack is a serious operating scaffold for a high-judgment communications lead who needs the leverage of a compact team without outsourcing editorial judgment to automation.

This repo is built around one principle:

AI prepares. Humans decide.

It is not a chatbot. It is not a press-release generator. It is a structured system for:

• daily market and media sensing
• stakeholder preparation
• campaign design and execution support
• founder-story packaging
• data-franchise development
• local-global coordination
• crisis readiness
• institutional memory compounding

Why this exists

Communications work in Ireland is dense with context, timing, relationships, and political nuance. Too much of that usually lives in inboxes, ad hoc notes, and people's heads.

Comms-Stack externalizes that operating intelligence into a repo-native system made of:

• a persistent knowledge/ layer
• reusable templates/
• specialist project-local skills in .cursor/skills/
• explicit operating rules in CLAUDE.md
• a full design spec in SYSTEM_SPEC.md

The repo is designed so a strong operator can move faster, stay sharper, and leave behind a system that compounds rather than disappearing with them.

Core ideas

1. The agents are the interface. The knowledge layer is the moat.
2. Preparation is more valuable than pseudo-autonomous output.
3. The best comms system improves judgment instead of replacing it.
4. Small-market relationship intelligence is strategic infrastructure.

Active command set

Daily heartbeat

• /landscape — daily market, media, competitor, and policy briefing
• /brief — one-page stakeholder preparation pack
• /rolodex — relationship memory update after meaningful interactions

Campaign surge

• /narrative — argument stack and thesis design
• /angles — audience- and outlet-specific hooks
• /draft — working drafts aligned to voice memory
• /campaign — sequencing, assets, approvals, and dependencies
• /data — aggregate-data franchise and evidence packaging
• /founder-voice — founder-story pipeline and prep
• /linkedin — executive and amplification posts in the right voice
• /monitor — post-launch pull-through and correction tracking
• /retro — system-improving retrospectives

Coordination

• /sync — SF-Dublin timing, translation, and escalation support
• /policy — comms-relevant Irish and EU policy interpretation
• /internal — employee-facing translation of external moments and incidents

Emergency

• /warroom — crisis structure, holding posture options, and stakeholder maps

Repo structure

.
├── .cursor/skills/        # project-local skills for all active agents
├── knowledge/             # narrative, relationship, policy, crisis, and editorial memory
├── templates/             # reusable output templates and YAML record shells
├── docs/                  # shared operating guidance
├── CLAUDE.md              # repo-local agent operating rules
├── SYSTEM_SPEC.md         # canonical design specification
├── PROJECT_TECHNICAL_OVERVIEW.md
├── CHANGES.md
└── workflow_state.md

How to use it

1. Learn the architecture

Read:

• SYSTEM_SPEC.md
• docs/OPERATING_MODEL.md
• CLAUDE.md

2. Use the knowledge layer as the system of record

Start from:

• knowledge/stripe-narrative.md
• knowledge/ireland-media-graph.md
• knowledge/policy-landscape.md
• knowledge/voice-memory.md
• knowledge/crisis-playbooks.md

3. Run the system at two speeds

Daily heartbeat:

1. /landscape
2. /brief
3. /rolodex

Campaign surge:

1. /narrative
2. /angles
3. /campaign
4. /draft
5. /brief
6. /linkedin, /founder-voice, and/or /data
7. /monitor
8. /retro

4. Use the templates instead of improvising structure

Key templates:

• templates/brief-template.md
• templates/campaign-template.md
• templates/incident-template.md
• templates/retro-template.md
• templates/data-report-template.md

Skill design

Each skill directory contains:

• SKILL.md — concise activation and workflow instructions
• reference.md — deeper strategic guidance, decision lenses, and failure modes

This keeps the main skill discoverable and tight while preserving enough depth to operate well under pressure.

Current maturity

This is a high-quality scaffold, not a finished product. It already includes:

• a complete system spec
• a seeded knowledge tree
• reusable templates
• 16 advanced project-local skills
• explicit operator rules for safe usage

What it does not include yet:

• live monitoring integrations
• automated data pipelines
• graph-backed relationship storage
• production orchestration

Operating boundary

This repo is intentionally strict about one thing:

No outbound autonomy.

AI can analyze, prepare, compare, package, and draft. Human operators retain the final call on outreach, publication, quotes, crisis posture, and sensitive messaging.