AI Agent Workflow Collection

Scaffolds an Architecture Decision Record (ADR-NNN.md) at a decision moment — allocates the next global-monotonic number across BOTH core/ADRs/ and release/ADRs/, stamps the derivable metadata (title stub, status Proposed, today's date, release slug, canonical Nygard section headers), and leaves every decision-prose section an author-fill placeholder. Scaffold-only: writes what is DERIVABLE, never the rationale that must be DECIDED (No-invention). Fires on an explicit request; a non-blocking passive offer may surface for an ADR-threshold decision but never auto-writes. Reads the ADR home DYNAMICALLY (never hardcoded). Immutable: allocates the next free number, never reuses or renumbers; supersession is a Status transition plus a new ADR. Consumes adr-authoring-guide.md + adr-schema.md; composes decision-discipline.md by reference. Triggers: "record this as an ADR", "scaffold an ADR", "create an ADR", "start an ADR", "write up this decision as an ADR", "adr-helper", "allocate the next ADR number".

Produces new or updated project artifacts — triggered by user request, PPM Agent gap detection, or phase gate requirements. Stages all output in 08-Generated/ with metadata for user review before promotion.

Scans a project's generated-artifact surface (08-Generated/ + promoted folders) and runs five graph-integrity checks — orphan, sibling duplicate, stale draft, displaced content, version chain — reading the horizontal-lineage frontmatter (parentartifact / siblingtopic / supersedes) so duplicate, orphaned, stale, or displaced generated content is surfaced for operator approval instead of silently accumulating. Read-and-recommend only — no automatic file moves or deletes.

Conducts final production-readiness review of any governed document pack — Copilot Builder Agent, PMO platform, or generic — using the shared review discipline (see core/disciplines/review-discipline-principles.md). Domain-specific dimensions loaded from pluggable dimension packs under references/dimension-packs/ (3 initial packs: copilot-builder, pmo-platform, generic). Produces a structured findings register with severity, affected documents, root cause, evidence, resolution, and reversibility tier per finding — consumed by implementation-planner downstream. Use when the user wants a final production-readiness review of a document pack, including: "review the copilot builder files", "audit the pmo platform", "review this document pack", "find gaps in the builder doc pack".

The voice of the PMO — produces audience-calibrated, ready-to-send communications. Covers email, Teams, Confluence, exec briefs, meeting agendas, escalation drafts, recaps, and status updates. Use when drafting any stakeholder communication.

Measures token consumption across every loadable pmo-platform component in a Claude Code / Cowork session — the SKILL.md catalog, core/rules/, core/standards/schemas/specs/disciplines, agents, hooks, and the CLAUDE.md context-file chain — and reports total + per-component estimated token cost, a flagged-bloat section with the applied thresholds stated inline, and a K1/K2 load-vs-reference hint per component. Single Measure mode. Estimates are a heuristic, trend-over-time signal (not falsely-precise absolutes), and the skill roster is sourced live, never hardcoded. Read-and-report only — measures cost, changes nothing. Triggers: "audit the context budget", "measure token consumption", "how much context does the platform load", "what's the per-component token cost", "flag context bloat", "which components are bloated", "context budget report".

Generates Teams-ready AM and PM daily status updates from carry-forward trackers and recent transcripts. Uses the project's Daily Status Update Framework.

Operational backbone for backlog health through release readiness. Modes: Backlog scan · Ticket insight · DoR gate · Sprint planning · Execution control · DoD gate · RAID updates. Use for sprint planning, backlog review, quality gates, or velocity tracking across Agile and Waterfall governance.

Authors rigorous eval suites for AI agents, skills, and LLM systems — grounded in the 2026 eval-writing consensus (trace-driven error analysis, binary LLM judges, cross-family validation, α/κ agreement). Produces characterization, failure taxonomies, judge prompts, rubrics, and calibration protocols that harnesses (pmo-skill-refiner, CI) then execute. Two modes — Author (write from scratch) and Review (audit against the framework). First-class playbooks for per-skill evals and for pipeline stage-gate judgment content; generic fallback for arbitrary AI systems. Use whenever the user asks to write evals, audit evals, add eval coverage, calibrate a judge, build a rubric, write a judge prompt, or diagnose why a judge keeps passing broken outputs.

Classifies, routes, and triggers processing for new files arriving in the PMO workspace. Uses three-layer classification (content analysis, project identification, filename patterns) with confidence thresholds.

Extracts per-session and subagent token-spend from local Claude Code sessions into the operator- local, git-ignored FinOps usage store, rolls it up to its work item, reports windowed spend and trends by work item, worktree, skill, MCP and model, and ESTIMATES a planned item's token cost from historical comparables — with a stated basis and confidence, declining below three comparables rather than implying false precision. Read-only; every figure carries its provenance and prints to stdout. Distinct from context-budget-auditor, which measures STATIC corpus footprint; this measures RUNTIME session spend. Triggers: "extract token usage", "refresh the finops store", "how much token spend", "run the finops extractor", "roll up token usage", "attribute token spend", "report token spend", "finops report", "spend by work item", "token spend trend", "where did the tokens go", "estimate token cost", "how many tokens will this take", "finops estimate", "token budget for this work item", "estimate vs actual token spend".

Scans a project's generated-artifact surface (08-Generated/ + promoted folders), groups retirement candidates into three groups (promoted-and-stale, approaching-timeout, superseded), and stages a cleanup proposal for unconditional operator approval before any archive action. Keys on the reconciled lifecycle fields (lifecyclestate / promotionstate / lifecyclechanged) plus a derived staleness signal and the artifact-lint report; never reads the deprecated single-field workflow machine. The archive action branches by promotionstate: a staged file is swept to archived/ (archived-in-place); a promoted file is content-retired in place (no move). Recommend-only at Autonomy Tier 1 — nothing is moved or archived without approval; a scheduled run stages a proposal that never self-applies. Distinct from cleanup-orphan-state.sh (a state-file tool, never invoked here). Triggers: "clean up 08-Generated", "run generated cleanup", "archive stale generated artifacts", "what generated files can be retired".

The conversational front door for intake — turns a half-formed idea into a well-formed, correctly-typed, correctly-placed work item logged to the work tracker. Meets you at any altitude (a single bug or a portfolio initiative), proposes the work-item type and its place in the intake hierarchy, elicits the type- and level-appropriate fields (a bug's reproduction and environment; a story's acceptance criteria and value), applies the 5-test rule live, confirms the drafted item with you, then logs it — never a scratch file. One work item per request (child candidates are noted in the body for later slicing; it does not auto-decompose). Two modes: Elicit (guided interview) and Triage-readiness check (run the 5-test against a draft you already wrote). Use when the user says "help me file this idea as an issue", "turn this into a work item", "log this idea", "what type of work item is this", "scope this idea for intake", "is this intake-ready", or "help me write up this bug/story/initiative".

System-scope Architect Specialist — cross-component, middleware, and data architecture; integration design and blast-radius assessment; system-level ADR authorship. Composes pmo-principal-engineer (within-component solution depth) and pmo-technical-analyst (technical review) — invokes each, re-implements neither. Modes: System-Design · Integration-Review. Use for system design across components, integration design, blast-radius assessment, cross-component architecture, or to write a system ADR. Triggers: "architecture review", "system design", "integration design", "blast-radius assessment", "cross-component design", "write a system ADR". Distinct from the solution-scope Principal Engineer (within-component depth).

Business Analyst Specialist — owns the HOW-it-works of delivery: requirements elicitation, process/workflow documentation, traceability, and gap analysis. Operates at the requirements tier: eliciting what the business needs, documenting how work flows, and proving the requirement traces from source to evidence. Composes pmo-process-designer (requirements definition, INVEST, Given-When-Then, process docs, traceability, gap analysis) + delivery-engine (backlog substrate, DoR readiness) — invokes them, never re-implements them. Modes: Requirements Elicitation · Workflow & Process Documentation · Traceability & Coverage · Gap Analysis. Use when the question is what the business needs elicited, how a process is documented, whether a requirement traces end-to-end, or where the coverage gap is. Triggers: "elicit the requirements", "document this process/workflow", "build the traceability matrix", "trace this requirement to Jira", "what's the gap", "map the as-is/to-be", "is this requirement complete".

DevOps/SRE Specialist — owns deploy mechanics + reliability/rollback triggers; composes release-executor — invokes it, never re-implements it. Executes the go/no-go the operator/Release-Manager already made; never makes the go/no-go itself. Modes: Pipeline · Deploy-Exec · Reliability. Use when a release needs its deploy run, its pipeline/rollout configured, or a reliability signal turned into a rollback proposal. Triggers: "configure the deploy pipeline", "run the deploy", "push the release through the pipeline", "set up the rollout", "we have a regression — trigger the rollback", "is the deploy healthy".

Knowledge Manager Specialist — captures, structures, routes, and stewards project knowledge assets (lessons learned, runbooks, decision records, reference docs) so they land in their governed home, not as two disconnected generate-then-route steps. Composes artifact-generator (produces/stages the asset) + file-router (classifies + routes it to its governed home) — invokes them via the core/ registry skill-chain, never re-implements them (ADR-019). Modes: Capture · Structure/Steward · Route/File · Knowledge-Gap Audit. Use when you want the platform to act as the Knowledge Manager. Triggers: "act as knowledge manager", "capture this as a knowledge asset", "steward the knowledge base", "file this lesson learned", "build the runbook and file it", "where does this knowledge go", "what knowledge are we missing".

Organizational Change Management (OCM) Lead Specialist — emulates an OCM Lead driving a go-live's organizational change end-to-end by composing the change-management function-skill across the full lifecycle. Composes change-management (impact · training · readiness · hypercare · adoption · comms) — invokes it through the core/ registry skill-chain, never re-implements it (ADR-019). Modes: Change Impact · Training & Adoption · Readiness Go/No-Go · Hypercare & Adoption Outcome · Change Comms Program. Use when the platform should act as the OCM lead, run or own the change program for a go-live, or drive organizational change as one coherent workflow. Triggers: "act as OCM lead", "run the change program", "own the change for this go-live", "drive organizational change", "lead the org change for this rollout", "manage the change end-to-end".

Portfolio Manager Specialist — synthesizes the cross-project health picture into portfolio-level calls. Operates at the portfolio tier above any single project, deciding which cross-project signal moves the portfolio posture and binding the roll-up's health verdicts to the portfolio's resource and capacity trade-offs. Composes ppm-agent (strategic triage + intake-governance) + weekly-status-rollup (cross-project roll-up + portfolio health write-back) — invokes them, never re-implements them. Modes: Portfolio Health Synthesis · Cross-Project Risk & Decision Triage · Portfolio Intake & Trade-off · SteerCo / Executive Readiness. Use when leadership needs the portfolio-altitude health, risk, intake, or SteerCo-readiness call across projects. Triggers: "what's the portfolio health", "what cross-project risk needs my attention", "triage the portfolio", "should the portfolio take this on", "score this new initiative", "prep the SteerCo", "portfolio summary for leadership".

Principal Engineer Specialist — solution-scope design depth at Stage 5. Decides the architecture for a within-component solution, governs its NFRs, adjudicates build-vs-buy, and authors the ADR when the decision is non-obvious. Composes pmo-technical-analyst (technical review) — invokes it, never re-implements it. Modes: Architecture & NFR Governance · Build-vs-Buy & Design Review. Use for solution-level design decisions distinct from system-scope architecture. Triggers: "design this solution", "what NFRs bind this design", "build-vs-buy for this component", "architecture decision for this solution", "is this design implementation-ready", "author the ADR for this decision".

Converts business context into structured, traceable requirements and process documentation. Modes: Requirements definition · Workflow documentation · Gap analysis · Traceability matrix · Compliance mapping. Use when uploading business requirements, FDDs, or Jira exports.

Product Owner Specialist — owns the WHAT and the WHY of the backlog: value framing, prioritization, story authoring with acceptance criteria, and the accept/reject decision. Operates at the backlog tier, deciding which item is most valuable to build next and whether its value is accepted. Composes pmo-process-designer (story/requirement authoring, INVEST scoring, Given-When-Then acceptance criteria) + delivery-engine (backlog health, DoR readiness, sprint-scope fit) — invokes them, never re-implements them. Modes: Backlog Prioritization · Story Authoring & Acceptance · Backlog Readiness · Value & Scope Decision. Use when the question is what to build next, how to value-rank the backlog, whether a story's acceptance criteria are accepted, or whether to include / defer / cut scope by value. Triggers: "prioritize the backlog", "what's most valuable", "rank these stories", "is this ready to pull", "should we build or defer X", "write the story and acceptance criteria for [value]", "accept this story".

Program Coordinator Specialist — keeps a program's operational trackers and status cadence coherent. Operates at the project tier across the program's workstreams, reconciling tracker state against the status cadence so nothing falls through the daily rhythm. Composes tracker-manager (tracker writes) + daily-status (AM/PM/Connect cadence) — invokes them, never re-implements them. Modes: Cadence Sync · Tracker Reconciliation. Use when the program's trackers and status updates need to be kept in lockstep. Triggers: "keep the trackers and status in sync", "reconcile the trackers before the update", "run the cadence sync", "did anything drop between the tracker and the status", "coordinate the program's daily rhythm", "tracker-to-status coherence check".

Program Manager Specialist — drives a program's multiple coordinated workstreams to outcomes, deciding which cross-project signal moves the delivery plan and binding the program's risk posture to its delivery decisions. Operates at the program tier above any single project's delivery and below portfolio strategy, across agile and waterfall governance. Composes ppm-agent (strategic synthesis + intake-governance) + delivery-engine (backlog→release-readiness mechanics) — invokes them, never re-implements them. Modes: Cross-Project Risk & Dependency Synthesis · Program Delivery Posture · Program Release-Readiness · Program Intake & Capacity Trade-off · Program RAID & Decision Stewardship. Use when one program's multi-workstream delivery needs a risk read, a posture call, a go/no-go, an intake trade-off, or RAID stewardship. Triggers: "program-level risk to delivery", "is this program ready to release", "across these workstreams", "program go/no-go", "cross-project dependency", "how's the program tracking".