Delta Report · Sandeep Swadia · Jul 30 2026 · 21 min

Four C's agents: you already run three and a half of them

Coordination, Creativity, Clarity, Coaching. Mapped against the Donna stack, with the two pieces actually worth stealing.

HAVEThe four pillars vs your stack

Pillar	His version	Your version	Status
Coordination	Gmail + Calendar agent, morning scheduled run, staged trust (read, then draft, then schedule)	Karen zero-inbox (9am/5pm), morning brief, 7AM digest, TickTick, Gcal-before-schedule rule	COVERED, deeper
Creativity	Folder-pointed deck building, brand template turned into a reusable skill	Claude Code + skills (docx/pptx/pdf), dwad-design system as a locked skill, content-tap pipeline	COVERED
Clarity	Telescope (multi-source pull) + microscope (contract teardown table), cross-model advisory board	Athena SOTA protocol, /td teardown, /spar, research-wiki. Contract-table format itself: not codified	90%, one steal
Coaching	Voice-mode rehearsal with persona interviewer, break-character debrief, tougher-persona rounds, one-page prep card	EO Forum prep flow exists. No rehearsal loop, no voice sparring partner, no debrief-then-harder-round cycle	REAL GAP

His meta-lessons are already Donna canon: staged trust (his "agents earn promotion" = your Angels' leash), clarity-in-equals-clarity-out, boundaries in every prompt ("draft, never send" = your escalation rules).

NEWWorth stealing

1. The Coaching agent, the only genuine gap. Persona-based voice rehearsal: agent plays a specific person (skeptical board member, tough buyer), one question at a time, pushes back on weak answers, then breaks character and coaches, then re-runs harder. His detail that lands for you: CEOs rehearsing board meetings with one agent per board member.
2. The microscope contract table. Not "summarize this contract" but a 5-column extraction: what it says, plain English, why it matters, risk level, questions to ask. A prompt shape, cheap to codify.
3. The 5-part prompt skeleton (job, tool, categories, output, boundary) is a decent teaching frame for the fCAIO/LMS audience even though your own prompts already do this implicitly. Content fodder more than tooling.

ACTImplementation plan (your stack, not his)

Build skills/rehearse.md, a /rehearse skill. Intake via AskUserQuestion (scenario, counterpart persona, stakes, duration), then run the loop: in-character interrogation one question at a time, push back on weak answers, break character on "done", deliver a graded debrief, offer a harder round, end with a one-page prep card saved next to the event. Voice happens on the Claude mobile app against the same skill, zero new infrastructure. First live use: EO Forum monthly prep and any Pasig broker or buyer negotiation.
Add the microscope table to skills/teardown.md as a document mode: when the target is a contract or policy, output the 5-column table instead of the 8-lens spar. One section, ten lines. Pasig broker sheets and the VGL/LMS engagement docs are the obvious first targets.
Skip everything else. Coordination and creativity agents: already outbuilt. His Coda scheduling: you have cron + scheduled tasks. Advisory board of 3 models: you already cross-verify via grunt-work routing when it matters.

Cost of the whole delta: one new skill file + one section edit. No new tools, no new subscriptions, leash intact.

NOISESkippable

Beginner walkthroughs of connecting Gmail connectors, the borrowed-tie interview story, Chaplin's Modern Times closer, job-displacement survey stats. Zero loss.