paivot-graph

Capture UX/visual/functional feedback and turn it into a prioritized backlog of high-quality stories using the Sr. PM agent

Cancel active execution loop

Run unattended execution loop until blocked or all work is done

Capture knowledge from the current session to the vault with auto-tagging and link suggestions

Refine vault knowledge based on session experience. Capture learned patterns, decisions, and debug insights. Agent prompts are self-contained in agents/*.md (not vault). System-scoped notes get proposals; project-scoped notes get direct edits.

Use this agent for adversarial review in TWO modes. (1) BACKLOG REVIEW (default) - Review backlog for gaps, missing walking skeletons, horizontal layers, missing integration stories. Must approve before execution. (2) MILESTONE REVIEW - After milestone completion, validate real delivery, inspect tests for mocks (forbidden), verify skills were consulted. Returns VALIDATED or GAPS_FOUND. Examples: <example>Context: Sr. PM has created the initial backlog from D&F docs. user: 'Review this backlog for gaps' assistant: 'I'll engage the anchor to adversarially review the backlog.' <commentary>Default mode - backlog review.</commentary></example>

Use this agent to adversarially review ARCHITECTURE.md after the Architect produces it. Only spawned when dnf.specialist_review is enabled. Reviews for unmet requirements, untraceable decisions, and contradictions across BUSINESS.md and DESIGN.md. Returns REVIEW_RESULT (APPROVED/REJECTED). Never talks to user -- feedback routes to Architect via dispatcher. Examples: <example>Context: Architect has completed ARCHITECTURE.md and specialist review is enabled. user: 'Review ARCHITECTURE.md for alignment with BUSINESS.md and DESIGN.md' assistant: 'REVIEW_RESULT: REJECTED with 2 issues (OMISSION: no authentication architecture for HIPAA constraint in BUSINESS.md, DRIFT: async pattern contradicts real-time requirement in DESIGN.md). Feedback provided for Architect to fix.' <commentary>Architect Challenger catches technical gaps before they cascade into the backlog and execution.</commentary></example>

Use this agent when you need to design system architecture, validate technical feasibility, or maintain architectural documentation. Part of the Balanced Leadership Team that communicates with the user through the orchestrator. Asks clarifying questions about technical constraints, existing infrastructure, team capabilities, and non-functional requirements. Owns ARCHITECTURE.md and ensures technical coherence across the system. Examples: <example>Context: Business Analyst presents new requirements that need technical validation. user: 'The BA says we need real-time data updates with 1-second latency for 50,000 concurrent users' assistant: 'I'll engage the architect to assess technical feasibility. I will relay its questions to you and pass your answers back until ARCHITECTURE.md is complete.' <commentary>The Architect will ask about existing infrastructure, deployment targets, budget constraints, and team experience before making decisions.</commentary></example> <example>Context: BLT cross-review after all D&F documents produced. user: 'Cross-review BUSINESS.md and DESIGN.md for consistency with ARCHITECTURE.md' assistant: 'I'll engage the architect to verify that business constraints and design patterns are technically feasible and properly reflected in the architecture.' <commentary>Architect reviews other BLT documents for technical consistency.</commentary></example>

Use this agent to adversarially review BUSINESS.md after the BA produces it. Only spawned when dnf.specialist_review is enabled. Reviews for omissions, hallucinations, and scope creep against user-provided context. Returns REVIEW_RESULT (APPROVED/REJECTED). Never talks to user -- feedback routes to BA via dispatcher. Examples: <example>Context: BA has completed BUSINESS.md and specialist review is enabled. user: 'Review BUSINESS.md for completeness against the user requirements' assistant: 'REVIEW_RESULT: REJECTED with 2 issues (1 OMISSION, 1 HALLUCINATION). Feedback provided for BA to fix.' <commentary>BA Challenger catches document-level issues before they cascade into DESIGN.md and ARCHITECTURE.md.</commentary></example>

Use this agent when you need to understand business requirements during Discovery & Framing. Part of the Balanced Leadership Team that communicates with the user through the orchestrator. Asks multiple rounds of clarifying questions until fully satisfied. Owns BUSINESS.md. Examples: <example>Context: User describes a business need for a greenfield project. user: 'We need to add authentication to our application' assistant: 'I'll engage the business-analyst to conduct thorough discovery, asking multiple rounds of clarifying questions. I will relay its questions to you and pass your answers back until BUSINESS.md is complete.' <commentary>The BA will dig deep through multiple questioning rounds until all ambiguities are resolved.</commentary></example> <example>Context: BLT cross-review after all D&F documents produced. user: 'Cross-review DESIGN.md and ARCHITECTURE.md for consistency with BUSINESS.md' assistant: 'I'll engage the business-analyst to check that business outcomes and constraints are properly reflected in the design and architecture.' <commentary>BA reviews other BLT documents for alignment with business requirements.</commentary></example>

Architecture-as-code using C4 model and Structurizr DSL. Use when the project has `architecture.c4` enabled in settings, or when the user asks about C4 diagrams, Structurizr, architecture boundaries, or dependency rules. Teaches agents how to maintain a canonical architecture model alongside ARCHITECTURE.md, declare machine-checkable boundaries, and export diagrams.

This skill should be used when working on any project to understand how to effectively interact with the Obsidian knowledge vault. It teaches when to capture knowledge, what to capture, how to format vault notes, and how to search effectively. Use when you need to "save to vault", "update vault", "capture a decision", "record a pattern", "log a debug insight", or when starting/ending a significant work session.

10 hooks across 9 events