โšก Valent Kickoff Prep Kit

๐Ÿ”’ INTERNAL โ€” SOW ref: VTKL-SOW-2026-VP-001 | Kickoff: June 10, 2026
๐Ÿ“‹ Kickoff Prep Kit๐Ÿ”ง Tactical / Operational๐ŸŽฏ Strategic

1. Kickoff Agenda โ€” 45 Minutes

Attendees: Tony (VTKL, intro), Charlie (VTKL, CTO), Joana (VTKL, lead & client touchpoint), Victor (VTKL, background) | Hector, Nick, RJ, Matt (Valent) | Charlie, Tony (TBD)

TimeTopicLeadNotes
0:00โ€“0:05Intro & RelationshipTonySets the tone. Introduces team: Tony (GTM), Charlie (CTO), Joana (delivery lead & primary contact). "Exploration, not a pitch."
0:05โ€“0:15Pilot Scope + What Warren DeliversCharlie / JoanaCharlie walks through capabilities (CTO credibility). Joana connects to Valent's delivery context. 4 deliverables, customizable.
0:15โ€“0:25What We Need From YouJoanaProject docs, Teams admin, channel. "Upload to Teams โ€” Warren reads automatically."
0:25โ€“0:35Working Model & Success CriteriaJoanaCadence, Joana in channel as support. "Minor edits, not rewrites." Let them add criteria.
0:35โ€“0:40Q&AAllTechnical โ†’ Charlie/Victor. Delivery โ†’ Joana/Victor. Strategy โ†’ Tony.
0:40โ€“0:45Next StepsJoanaAction items, owners, Day 1 of 14-day timeline.

Facilitator notes: Keep Hector on strategic value. Let RJ drive "how will I use this daily" questions. Answer Matt's technical questions directly and honestly. Frame: "our team configures and maintains this for you."

2. Persona Playbook

Hector Martinez โ€” Partner

Background: 35+ years consulting/telecom. CS degrees (SMU, Baylor). Former CSO at Pariveda. Sprint, Level 3, Accenture.

Cares about: Customization & flexibility. Strategic value to Valent's practice. Scale across client engagements. Revenue impact (1-2 FTE equivalent = $200K-$500K/yr).

Wants to hear: "You can change format, detail level, graphics, color. It adapts to each client's standards." "Real SDLC potential."

Avoid: Over-technical details. Rigid product framing. Anything implying Valent loses control of client relationships.

Key signal: "That's money" (on SDLC). "How easy to change roadmap format?"

Nick Niver โ€” Partner, Strategy

Background: Leads Business Transformation. Strategy-oriented, big-picture.

Cares about: How this fits Valent's strategy practice. Competitive differentiation. Scale potential.

Wants to hear: "Augments your consultants โ€” doesn't replace them." "Spin up an AI partner for every new engagement within days."

Avoid: Implying AI replaces consulting judgment. Cost-cutting framing vs capability multiplier.

RJ Beecher โ€” Managing Consultant, PMO

Background: Hands-on PMO practitioner. Primary daily user. His experience determines expansion.

Cares about: "Can I use this?" Day-to-day workflow. Quality. Time saved vs time learning.

Wants to hear: "Talk to it like a colleague in Teams. Ask for a status report, it produces one."

Avoid: Abstract strategy talk. Jargon. Making him feel like a test subject.

Matt Malone โ€” Principal, AI

Background: Technical evaluator. Assesses AI capability and quality.

Cares about: AI quality/accuracy. Hallucination risk. Data handling. Architecture (high level).

Wants to hear: "Frontier models. The value is agent configuration + methodology encoding, not just the model." Be honest about limitations.

Avoid: Hand-waving. Claiming perfection. Exposing internal tooling.

3. Data Collection Checklist

#ItemPriorityNotes
1Teams/M365 admin contactBLOCKINGName + email for app installation
2Approval to install Teams bot appBLOCKINGMay require IT security review
3Target team/channelBLOCKINGDedicated channel recommended
4Historical project documentationBLOCKINGDeliverables, meeting notes, artifacts through ~midpoint (3-5 person project)
5PMO methodology templatesHIGHStatus report format, sprint report format โ€” customize to their standards
6Format preferencesHIGHWord templates, branding, color, logo
7Security/compliance constraintsHIGHData handling, cloud hosting reqs, BAA needs
8Communication cadence preferenceMEDIUMWeekly sync? Async-first?
9Escalation pathMEDIUMWho to contact for issues
10Jira/Confluence access statusLOW (Phase 2)Future planning only โ€” Phase 1 is doc-based

Talking point: "We need enough project context to simulate being dropped into an active engagement at the midpoint."

4. Internal Risk Register

1. Bootstrap Gap HIGH

Risk: 5 days to build what took 3+ months with the production system. Clean install = no memory, no context, no calibration.

Impact: Pilot outputs underwhelm, Valent loses confidence.

Mitigation: Production system coaches behind scenes (GitHub architecture). Pre-seed memory with Valent context and AIPMO methodology. Front-load configuration before kickoff.

Owner: Charlie

2. Data Control HIGH

Risk: If deployed on client infrastructure, we lose access to logs, outputs, iteration ability.

Impact: Can't iterate, debug, or improve quality during pilot.

Mitigation: All infra on VTKL AWS. Talking point: "Dedicated secure environment managed by our team for optimal quality."

Owner: Charlie

3. Quality Calibration MEDIUM

Risk: System calibrated for agentic engineers; Valent has human teams.

Impact: Deliverables miss the mark on tone, detail level, or methodology.

Mitigation: Seed with PMO-specific methodology. Phase 1 = deliverables only (not estimation). Gather calibration data during pilot.

Owner: Victor

4. Deterministic vs Probabilistic MEDIUM

Risk: Paradigm shift in progress internally. Risk of quality drop without intermediate gates.

Impact: Inconsistent output quality.

Mitigation: Client deployment keeps analytical/process-following tier. Internal system experiments with new paradigm separately.

Owner: Tony

5. Naming/Branding Confusion LOW likelihood / HIGH impact

Risk: Internal and external names differ. Client sees inconsistency.

Mitigation: Decide before kickoff: one name. Recommendation: "Warren" (matches signed SOW).

Owner: Victor

6. Jira Complexity MEDIUM

Risk: Valent uses CLIENT Jira, not their own. BAAs, healthcare data complications.

Mitigation: Phase 1 = non-Jira, doc-based only. Jira deferred to Phase 2 if Phase 1 succeeds.

Owner: Charlie

7. IP Exposure MEDIUM โ†’ MITIGATED (Jun 10)

Risk: Internal tooling, pipeline, other client details leak through agent responses.

Original mitigation: Sanitization rules in agent configuration.

โœ… Resolution: Workspace isolation implemented โ€” see Section 7. Parametrized Dockerfile defaults to clean base for client builds. Scope guardrail enforces deliverable whitelist. 106/131 contaminated memory files eliminated from client images. PR #14 โ€” merged.

Owner: Victor + Warren

8. Estimation for Human Teams LOW (Phase 1)

Risk: System calibrated for agentic engineers, not human teams.

Mitigation: Phase 1 doesn't include estimation. Defer to Phase 2 with methodology calibration.

Owner: Tony

5. Decisions for Charlie โ€” Monday Session

#DecisionContextRecommendation
1Architecture: GitHub-based config sync vs standalone?Golden repo โ†’ deploy model confirmed, but coaching mechanism undefined.GitHub config sync. Define mechanism: shared memory files via git, or manual knowledge transfer?
2NamingSOW says "Warren." Internal codename is different."Warren" for client-facing (matches signed SOW).
3Teams integration pathInstall bot in Valent's tenant vs invite to VTKL channel.Install in their tenant (matches SOW ยง2.1).
4Governance gates4 blocking items on golden repo (#6-#9). All truly blocking?SOUL.md (#6) + Operating Policies (#9) = true blockers. Scope (#7) + Client-Fit (#8) = parallel.
5Dogfood skip?Charlie's path was dogfood-then-Valent. Kickoff is June 10 โ€” 4 days.Likely no time. Options: fast dogfood Jun 6-9, skip, or delay kickoff.
6Config sync mechanismHow does the production system feed the client instance?Simplest: write memory/skill files to golden repo. Complex: real-time API.
7Who attends kickoff?VTKL confirmed: Victor, Joana.Charlie/Tony for intros (first 10 min) signals commitment.

7. Workspace Isolation โ€” Security Implementation (Jun 10)

Origin โ€” Why This Was Done

This work originated from the June 10, 2026 team calls (Tony, Charlie, Joana, Victor). Three converging signals triggered the implementation:

Victor brought the analysis to Warren, who audited the t-and-c-golden repo and confirmed: 106 of 131 memory files in the golden workspace contained references to Kindo, Deloitte, and other clients. The Dockerfile had no mechanism to exclude this from client builds. Implementation was executed same-day.

Status: PR #14 โ€” merged to main โœ…
Addresses: Risk #7 (IP Exposure), Joana's scope containment requirement, Tony's "instance factory" need

Problem Identified

The golden config base (workspace/) was a direct copy of Warren's full workspace:

Any client instance built from this image would carry all internal context.

What Was Implemented

1. Clean derivation base (workspace-base/)

2. Valent scope guardrail (Standing Rule 0)

3. Parametrized Dockerfile

# VTKL dogfood (unchanged behavior)
docker build .

# Client build (clean base + client overlay)
docker build --build-arg CLIENT=valent .

# Makefile shortcut
make client-build CLIENT=valent
make client-verify CLIENT=valent   # contamination check

When CLIENT is set, workspace auto-defaults to workspace-base (safe by default).

4. Deploy + onboard scripts wired with --client flag

5. Repo structure after isolation:

t-and-c-golden/
โ”œโ”€โ”€ workspace-base/     โ† CLEAN (client derivations)
โ”œโ”€โ”€ workspace/          โ† VTKL dogfood only (internal)
โ”œโ”€โ”€ clients/
โ”‚   โ”œโ”€โ”€ valent/workspace/   โ† scoped + isolated + guardrail
โ”‚   โ””โ”€โ”€ vtkl/workspace/     โ† dogfood overlay
โ””โ”€โ”€ config/

Verification

CheckResult
grep workspace-base for client names/strategyZero hits โœ…
Valent overlay for Kindo/Deloitte/RonZero hits โœ…
Assembly: default (no args)workspace (241 files) โ€” backwards compat โœ…
Assembly: CLIENT=valentworkspace-base + overlay (13 files, 0 contamination) โœ…
Scope guardrail positioned as Rule 0Above all other rules โœ…

What Still Needs to Happen

ItemOwnerStatus
Review + merge PR #14Victor / CharlieMerged โœ…
Red team validation (adversarial prompts)WarrenAfter merge
Response middleware (Level 2 enforcement)WarrenDeferred
Rebuild + redeploy Aria VTKLWarren + CharlieAfter merge

Why This Matters

Joana (Jun 10): The agent must guide the client through structured deliverable production, not answer open-ended questions. Exposing Warren's full context to Valent would leak Kindo, Deloitte, and client relationships.

Tony (Jun 10): Wants to "start breaking off pieces" of Warren for Kodazen, Valent, digital twins. This workspace isolation is the prerequisite for that instance factory.