AI Agent Workflow Collection

Validates ADR structural conformance for >: every ADR has the 5 required sections, every ADR is listed in INDEX.md, no orphaned ADRs exist, supersession links are bidirectional, and status fields are consistent between INDEX.md and ADR files.

Guides writing a new Architectural Decision Record in the canonical ASCENT format: Context, Decision, Alternatives considered, Consequences (Easier/Harder/Neutral), and Cost implications. Assigns the next ADR number, creates the file, and updates INDEX.md.

Validates SQLite-WAL database configuration and migration discipline for >. Checks WAL mode, foreign-key enforcement, migration tracking, and migration file ordering — the silent-failure prevention layer for the data tier.

Surfaces the current phase, outstanding work, exit-criteria progress, and recommended next actions for >. The delivery-lead's at-a-glance view of where the project is and what's needed to close the current phase gate. Per §15, incorporates session-state.md and working-memory.md into the synthesis when present.

Validates dependency discipline for > via static package.json analysis. Checks pinned versions, engine constraints, framework traceability, and unused dependencies — offline-capable, no network required.

Generates persona-targeted documentation skeletons for >. Each generated doc has a named audience, a clear navigation path from the persona's entry point, and section headings that match the audience's needs — not blank files or generic templates.

Validates documentation freshness and consistency for >. Checks persona provenance comments, orphaned references, stale TODO/TBD markers (30-day threshold), and doc-graph integrity per §14.

Verifies that > follows ENV-DISCIPLINE: .env is gitignored and dockerignored, .env.example has empty defaults only (no REPLACEME), every environment variable read in code appears in .env.example, and no secrets are committed or baked into images. Can run standalone or as a component of ascent-self-audit's umbrella.

Runs the delivery-lead intake protocol for new features in >: decomposes a request into acceptance criteria, identifies cross-role dependencies, estimates scope, and records the feature in the project's PHASE-PLAN. Per §15, appends locked acceptance criteria as dated entries to working-memory.md.

Documents >'s current state for another developer or session takeover. Produces a progressive-depth HANDOFF.md artifact at docs/delivery/HANDOFF.md covering current work, recent decisions, phase progress, and project conventions.

Verifies that >'s backend follows strict layering: routes → controllers → services → storage. Detects cross-layer violations (routes importing storage, controllers containing SQL, services using req/res) by analyzing import patterns and code structure. Can run standalone or as a component of ascent-self-audit's umbrella.

Proposes and creates new make targets for > following the MAKE-NAMING convention ( - - ), places them in the correct SDLC-sectioned .mk file, and validates the naming against existing targets to prevent aliases or collisions.

Verifies that >'s backend services emit observability data per the observability contract: structured JSON logs with the required fields, /healthz and /readyz endpoints, lifecycle events (service.started, service.draining, service.stopped), and trace-id propagation via W3C TraceContext. Can run standalone or as a component of ascent-self-audit's umbrella.

Guides a new developer through > with a structured "first hour" walkthrough. Read-only synthesis — detects project maturity from state and adapts depth accordingly. Covers identity, conventions, current state, key decisions, and where to start.

Surface-level quality gate for >. Runs lightweight checks across structural integrity, ADR discipline, dependency health, documentation freshness, and skill collection integrity. Conditionally recommends deeper sub-skills when concerns are flagged.

End-of-session capture for >. Interviews the developer for current focus, decisions made, blockers, and next actions, then writes to session-state.md (overwrite) and working-memory.md (append) per §15 session resumption protocol.

Pre-release validation gate for >. Checks CHANGELOG entry, version consistency, clean working tree, release branch, and tag uniqueness. Detects meta-repo vs scaffolded-project context and adapts version-field checks accordingly.

Verifies that > conforms to all 15 ASCENT invariants. The umbrella audit — delegates to ascent-layering-check, ascent-env-audit, and ascent-observability-check for their domains, then validates the remaining 12 invariants directly: containerization, make vocabulary, SDLC-sectioned help, stub-first naming, backend layering structure, ADR discipline, phase-gated delivery, label-based scoping, engine compatibility, context-aware execution, graceful shutdown, dev-status family, persona-segmented docs, and session resumption.

Validates the project-embedded skill collection integrity for >. Checks SKILL.md frontmatter, INTENT-MAP.md sync, cross-skill reference integrity, and allowed-tools honesty across all ascent- skills.

Generates a brief, time-scoped standup summary for >: recent activity (last 24 hours of git history + working-memory decisions), current focus, plan for next actions, and blockers. The standup reads the same project state as ascent-delivery-status but produces a concise output a developer reads in 30 seconds.

ASCENT engineering framework. Use to scaffold new projects, enhance existing ASCENT projects, migrate non-ASCENT projects to ASCENT standards, or answer questions about ASCENT's conventions, principles, or nine engineering roles (delivery-lead, architect, ui-ux-designer, developer, data-engineer, ai-engineer, tester, devops, cybersecurity).