pstack — Palantir Deployment Strategist AI Team

Turn one deployment strategist into a disciplined multi-agent Palantir delivery team.

pstack is an operator-grade repository of AI skills for Palantir deployments. It is built for deployment strategists who need an execution contract, not a loose collection of prompts: phase gates, artifact handoffs, governance boundaries, examples, evals, and a gated learning loop all live in one repo.

What This Is

pstack turns one deployment strategist into a conductor for discovery, ontology, data, application, QA, security, deployment, retrospective work, and repo improvement. The repo is organized around a strict artifact chain so downstream skills consume stable documents rather than improvising off chat history.

Who This Is For

• Deployment strategists running AIP bootcamps or compressed phase-zero engagements
• Foundry builders who need a disciplined artifact chain across ontology, pipeline, app, and QA work
• Operators managing multiple parallel AI sessions and wanting explicit phase gates

Who This Is Not For

• Teams looking for a generic prompt pack without phase discipline
• Pure software projects with no Foundry, ontology, or Apollo deployment surface
• Production changes without a human deployment strategist owning approvals

First Hour

1. Clone the repo and run ./install.sh.
2. Start with /bootcamp and produce BOOTCAMP-SCOPE.md.
3. Run /ontology-vision and /pipeline-plan in parallel once the use case is locked.
4. Promote only the headliner use case into /ontology-architect.
5. Do not start build skills until the architecture artifacts are reviewed.

First Week

1. Day 1: Lock the customer problem, use-case ranking, stakeholder map, and data inventory.
2. Day 2: Finalize ontology and pipeline architecture with explicit gating assumptions.
3. Day 3-4: Run build skills in parallel behind the artifact chain.
4. Day 4-5: Run review, security, and QA before any Apollo cutover discussion.
5. End of week: Produce deployment plan, training materials, and retrospective so the next phase starts from evidence instead of memory.

Quick Start

git clone https://github.com/myceldigital/pstack.git
cd pstack
./install.sh

# Start with structured discovery
/bootcamp
/ontology-vision
/pipeline-plan
/ontology-architect

Skills

The skill inventory below is generated from conductor.json.

Discover

Skill	Purpose	Reads	Produces
/bootcamp	Structured discovery for new engagements; produces the bootcamp scope and use-case priority.	Customer brief / engagement context	BOOTCAMP-SCOPE.md

Vision

Skill	Purpose	Reads	Produces
/ontology-vision	Expands the bootcamp scope into the full target digital twin and phased ontology vision.	BOOTCAMP-SCOPE.md	ONTOLOGY-VISION.md

Architecture

Skill	Purpose	Reads	Produces
/ontology-architect	Locks object types, links, actions, functions, interfaces, and ontology test plans.	BOOTCAMP-SCOPE.md, ONTOLOGY-VISION.md, PIPELINE-ARCHITECTURE.md	ONTOLOGY-ARCHITECTURE.md
/pipeline-plan	Designs source-to-ontology flow, connection strategy, transform DAGs, and scheduling.	BOOTCAMP-SCOPE.md, ONTOLOGY-VISION.md	PIPELINE-ARCHITECTURE.md

Build

Skill	Purpose	Reads	Produces
/data-connector	Configures data connections, raw datasets, first-stage cleaning, and health monitoring.	BOOTCAMP-SCOPE.md, PIPELINE-ARCHITECTURE.md	DATA-CONNECTION-STATUS.md
/pipeline-builder	Builds transforms, quality checks, and output datasets that back the ontology.	DATA-CONNECTION-STATUS.md, PIPELINE-ARCHITECTURE.md, ONTOLOGY-ARCHITECTURE.md	PIPELINE-BUILD-STATUS.md
/workshop-builder	Builds operational Workshop apps, wiring actions, variables, widgets, and flows.	BOOTCAMP-SCOPE.md, ONTOLOGY-ARCHITECTURE.md	WORKSHOP-BUILD-STATUS.md
/aip-architect	Designs agent prompts, context scope, tools, evals, and Automate workflows.	ONTOLOGY-ARCHITECTURE.md, WORKSHOP-BUILD-STATUS.md	AIP-AGENT-STATUS.md
/osdk-developer	Builds custom applications against the Ontology SDK when Workshop is insufficient.	ONTOLOGY-ARCHITECTURE.md	OSDK-BUILD-STATUS.md
/slate-builder	Builds Slate apps for complex, pixel-precise, or deeply integrated user experiences.	ONTOLOGY-ARCHITECTURE.md	SLATE-BUILD-STATUS.md

Review

Skill	Purpose	Reads	Produces
/foundry-reviewer	Performs structural audit of pipeline, ontology, app, and agent deliverables.	PIPELINE-BUILD-STATUS.md, WORKSHOP-BUILD-STATUS.md, AIP-AGENT-STATUS.md, OSDK-BUILD-STATUS.md, SLATE-BUILD-STATUS.md, ONTOLOGY-ARCHITECTURE.md	REVIEW-REPORT.md
/foundry-security	Audits permissions, markings, agent scoping, OAuth scopes, data governance, and deployment security.	ONTOLOGY-ARCHITECTURE.md, AIP-AGENT-STATUS.md, OSDK-BUILD-STATUS.md, PIPELINE-BUILD-STATUS.md, WORKSHOP-BUILD-STATUS.md	SECURITY-AUDIT.md

Test

Skill	Purpose	Reads	Produces
/foundry-qa	Runs end-to-end QA, evals, and readiness scoring before deployment.	REVIEW-REPORT.md, PIPELINE-BUILD-STATUS.md, WORKSHOP-BUILD-STATUS.md, AIP-AGENT-STATUS.md, SECURITY-AUDIT.md	QA-REPORT.md

Ship

Skill	Purpose	Reads	Produces
/apollo-deployer	Manages release channels, deployment strategy, rollback, and post-deploy verification.	QA-REPORT.md, SECURITY-AUDIT.md	DEPLOYMENT-PLAN.md
/training-writer	Produces customer guides, runbooks, decks, and enablement materials from the built system.	DEPLOYMENT-PLAN.md, QA-REPORT.md, SECURITY-AUDIT.md, ONTOLOGY-ARCHITECTURE.md, WORKSHOP-BUILD-STATUS.md, AIP-AGENT-STATUS.md	TRAINING-MATERIALS.md

Reflect

Skill	Purpose	Reads	Produces
/deployment-retro	Captures engagement metrics, friction points, lessons learned, and next-wave opportunities.	BOOTCAMP-SCOPE.md, ONTOLOGY-ARCHITECTURE.md, REVIEW-REPORT.md, QA-REPORT.md, DEPLOYMENT-PLAN.md, TRAINING-MATERIALS.md	RETRO-REPORT.md

Improve

Skill	Purpose	Reads	Produces
/memory-curator	Extracts reusable lessons from completed engagements into structured episodic and semantic memory.	BOOTCAMP-SCOPE.md, REVIEW-REPORT.md, QA-REPORT.md, RETRO-REPORT.md	MEMORY-EPISODE.json
/skill-improver	Turns structured memory into bounded repo-improvement proposals targeting skills, templates, PROGRAM.md, and conductor metadata.	MEMORY-EPISODE.json, REVIEW-REPORT.md, QA-REPORT.md, RETRO-REPORT.md	IMPROVEMENT-PROPOSAL.md

Safety

Skill	Purpose	Reads	Produces
/careful	Requires explicit DS confirmation before destructive Foundry operations.	Customer brief / engagement context	Policy mode
/freeze	Restricts all modifications to an explicit project, use-case, or environment scope.	Customer brief / engagement context	Policy mode
/guard	Combines careful and freeze for maximum safety in production customer work.	Customer brief / engagement context	Policy mode

Sprint Sequence

Phase	Goal	Primary skills	Primary artifacts
Discover	Lock the operational problem, customer context, and use-case priority.	/bootcamp	BOOTCAMP-SCOPE.md
Vision	Model the full digital twin and target ontology direction.	/ontology-vision	ONTOLOGY-VISION.md
Architecture	Turn the chosen use case into precise ontology and pipeline designs.	/ontology-architect, /pipeline-plan	ONTOLOGY-ARCHITECTURE.md, PIPELINE-ARCHITECTURE.md
Build	Stand up data connections, transforms, apps, and agents in parallel.	/data-connector, /pipeline-builder, /workshop-builder, /aip-architect, /osdk-developer, /slate-builder	DATA-CONNECTION-STATUS.md, PIPELINE-BUILD-STATUS.md, WORKSHOP-BUILD-STATUS.md, AIP-AGENT-STATUS.md, OSDK-BUILD-STATUS.md, SLATE-BUILD-STATUS.md
Review	Audit quality, security, and operability before testing.	/foundry-reviewer, /foundry-security	REVIEW-REPORT.md, SECURITY-AUDIT.md
Test	Run end-to-end validation and produce a deployment verdict.	/foundry-qa	QA-REPORT.md
Ship	Deploy safely and enable adoption.	/apollo-deployer, /training-writer	DEPLOYMENT-PLAN.md, TRAINING-MATERIALS.md
Reflect	Capture lessons, metrics, and next-phase opportunities.	/deployment-retro	RETRO-REPORT.md
Improve	Convert engagement evidence into reusable memory, improvement proposals, and gated repo evolution.	/memory-curator, /skill-improver	MEMORY-EPISODE.json, IMPROVEMENT-PROPOSAL.md

Artifact Chain

Every named artifact below has a template in templates/, rubric coverage in evals/ARTIFACT-RUBRICS.md, and live examples under examples/engagements/.

BOOTCAMP-SCOPE.md -> /ontology-vision, /pipeline-plan, /ontology-architect, /data-connector, /workshop-builder
ONTOLOGY-VISION.md -> /ontology-architect, /pipeline-plan
ONTOLOGY-ARCHITECTURE.md -> /pipeline-builder, /workshop-builder, /aip-architect, /osdk-developer, /slate-builder, /foundry-security, /foundry-qa
PIPELINE-ARCHITECTURE.md -> /data-connector, /ontology-architect, /pipeline-builder
DATA-CONNECTION-STATUS.md -> /pipeline-builder
PIPELINE-BUILD-STATUS.md -> /foundry-reviewer, /foundry-qa
WORKSHOP-BUILD-STATUS.md -> /aip-architect, /foundry-reviewer, /foundry-qa
AIP-AGENT-STATUS.md -> /foundry-reviewer, /foundry-security, /foundry-qa
OSDK-BUILD-STATUS.md -> /foundry-reviewer, /foundry-security, /foundry-qa
SLATE-BUILD-STATUS.md -> /foundry-reviewer, /foundry-qa
REVIEW-REPORT.md -> /foundry-qa, /deployment-retro
SECURITY-AUDIT.md -> /apollo-deployer, /deployment-retro
QA-REPORT.md -> /apollo-deployer, /deployment-retro
DEPLOYMENT-PLAN.md -> /training-writer, /deployment-retro
TRAINING-MATERIALS.md -> /deployment-retro
RETRO-REPORT.md -> end of chain
MEMORY-EPISODE.json -> /skill-improver
IMPROVEMENT-PROPOSAL.md -> end of chain

Operator Governance

Use PROGRAM.md as the field manual. It defines:

• when an artifact is good enough to advance
• what can run in parallel and what must block
• when to activate /careful, /freeze, or /guard
• what requires explicit DS approval
• how to stop, retry, or split an engagement when the artifact chain breaks

Memory, Examples, And Evals

• memory/ — structured episodic memory, semantic patterns, operator memory, and gated improvement proposals
• examples/engagements/acme-supply-chain/ — control-tower style supply chain engagement
• examples/engagements/northstar-healthcare/ — patient-flow style healthcare engagement
• evals/ — benchmark cases, artifact rubrics, and reference scorecards

Repo Structure

• PROGRAM.md: operator manual and execution contract
• conductor.json: source of truth for phases, skills, artifacts, governance, and examples
• memory/: self-improvement substrate for episodes, semantic memory, and proposal drafts
• docs/skills.md: generated detailed skill and artifact registry
• templates/: artifact templates for every conductor artifact
• examples/: end-to-end synthetic engagements
• evals/: benchmark cases, rubrics, and reference scorecards

Design Principles

See ETHOS.md for the operating philosophy. PROGRAM.md is the execution layer that turns those principles into actual gating, concurrency, and approval rules.

Requirements

• Claude Code (or equivalent environment that can consume the installed skills)
• Familiarity with Palantir Foundry, AIP, and deployment workflows
• A deployment strategist willing to own customer-facing decisions and approval gates

License

MIT