investiture

Survey an existing codebase and generate Investiture doctrine files — VECTOR.md, CLAUDE.md, ARCHITECTURE.md — by combining Investiture defaults with patterns inferred from the project. Run this once on a project that does not yet have doctrine. Then run invest-doctrine to validate what was generated.

Generate a complete product manifest of a codebase — every file, route, endpoint, database table, component, hook, and feature inventoried in a single structured document. Use when you need to understand what exists in a project before working on it.

Quick reconnaissance pass on a codebase to establish project type, tech stack, structure, scale, and conventions before running deeper skills like /invest-manifest or /invest-repo-audit. Run this first on any unfamiliar repo.

Read an AUDIT.md and generate a phased remediation plan — a sequence of self-contained, agent-executable prompts ordered from lowest-risk to highest-risk. Each phase is scoped to be completable in one session.

Perform a full quality, stability, and architecture audit of a codebase. Identifies broken features, dead code, spaghetti coupling, missing error handling, security gaps, performance problems, and technical debt. Produces a classified report with prioritized remediation.

After a remediation plan has been executed, verify that findings were resolved without introducing new issues. Updates AUDIT.md with resolution status, cleans up REMEDIATION.md, and confirms the codebase is ready to resume building.

Generates a numbered Architecture Decision Record from a decision description, using VECTOR.md constraints, ARCHITECTURE.md stack context, and existing ADRs to avoid contradictions. Use when making an architectural or product decision that should be documented and stored.

Generates a design brief for a specific feature or flow from existing project research. Reads personas, JTBD, VECTOR.md principles and constraints, and quality gates to give a designer actionable direction grounded in real evidence. Use before design work begins on a feature or flow.

Reads git log since the last tag and VECTOR.md value prop, groups commits by user-facing theme, filters internal noise, and writes a plain-language changelog keyed to what users can now do. Use at release time or when updating a public changelog.

Decomposes a feature into scoped agent tasks before a multi-agent sprint. Reads ARCHITECTURE.md layer ownership, CLAUDE.md agent model, and VECTOR.md constraints to generate a structured task manifest with branch names, commit prefixes, and explicit scope boundaries. Use before starting multi-agent feature work.

Generates a condensed onboarding document tailored to a specific role (engineer, designer, agent, client). Reads VECTOR.md, CLAUDE.md, ARCHITECTURE.md, and current /vector/ state. Use when onboarding a collaborator, starting a cold agent session, or preparing a client check-in.

Generates a structured user research discussion guide from unvalidated assumptions and open questions in /vector/. Produces targeted questions, probing techniques, and explicit validation signals per assumption. Use before any user research session.

Takes raw research input — interview notes, beta feedback, assumption validation results — extracts structured insights, and proposes specific patches to VECTOR.md and /vector/ schema files. Shows a full diff before writing anything. Use after user interviews or beta feedback rounds.

Reads /vector/research/assumptions/, prioritizes unvalidated assumptions by risk (Impact × Confidence), and generates a stage-appropriate validation plan. Use at the start of a sprint or before major feature investment to turn untested hypotheses into an actionable plan.

Survey an existing codebase and generate Investiture doctrine files — VECTOR.md, CLAUDE.md, ARCHITECTURE.md — by combining Investiture defaults with patterns inferred from the project. Run this once on a project that does not yet have doctrine. Then run invest-doctrine to validate what was generated.

Audit the project structure against ARCHITECTURE.md. Reads YOUR doctrine at runtime — layers, naming, stack, conventions — and checks the codebase against what YOU declared. Not a preset. Your rules, enforced.

Audit the doctrine files themselves — VECTOR.md, CLAUDE.md, ARCHITECTURE.md — for completeness, internal consistency, cross-document contradictions, and drift from the actual codebase. Run this before invest-architecture. The doctrine must be sound before you enforce it.