yeisme-auctra-cli-runtime

Develop and validate the Auctra CLI-first text creation runtime with review-safe evidence and typed handoffs.

Updated Aug 3, 2026
One-click install
npx skills add https://github.com/yeisme/auctra-novel-skills --skill yeisme-auctra-cli-runtime-yeisme
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: yeisme-auctra-cli-runtime
Source: https://github.com/yeisme/auctra-novel-skills/tree/main/auctra-runtime/yeisme-auctra-cli-runtime
Command: npx skills add https://github.com/yeisme/auctra-novel-skills --skill yeisme-auctra-cli-runtime-yeisme

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? It guides engineers changing, testing, reviewing, or documenting the Auctra CLI-first text creation product runtime under cli/auctra, enforcing product boundaries, human review invariants, and evidence integrity so changes do not break contracts. ## Core Features & Use Cases - Runtime Boundary Enforcement: Defines module ownership across internal/cli, internal/app, internal/content, internal/corpus, internal/review, internal/runtime, and internal/store, plus closed provider IDs (pi, omp, cohors, fixture). - Review-First Text Engine: Ensures generation only creates pending review items, with accepted versions produced solely through review accept or review partial across novels, Xiaohongshu notes, WeChat articles, and screenplay scenes. - Screenplay Corpus Intelligence & Scaena Handoff: Governs additive corpus overlays, holdout isolation, and the accepted-only auctra.production_handoff.v1 export contract with fail-closed readiness checks. - Use Case: When adding a new CLI command or changing run evidence output, follow this Skill to keep --json envelopes stable, redaction enabled, and run the required go test and openspec validate commands. ## Quick Start Ask the agent to modify or review a command in cli/auctra while preserving CLI ownership, human review, and evidence redaction, then run the specified Go tests and OpenSpec validation.

Frequently Asked Questions about yeisme-auctra-cli-runtime

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I add or change a command in the Auctra CLI?▼

Command parsing and mode wiring live in internal/cli and internal/cli/command, with workflows in internal/app. Agent-facing commands must support --json emitting one envelope or --agent emitting stable key=value facts, with diagnostics on stderr.

How does Auctra handle AI text generation review?▼

Generation only creates pending review items; accepted versions are produced exclusively through review accept or review partial. There is no --auto-accept, unattended overwrite, or publish bypass, keeping human review first.

What runtime providers does Auctra support?▼

Provider IDs are closed: pi is the default real runtime, omp is explicit via --runtime omp or AUCTRA_RUNTIME=omp, fixture is the only deterministic offline test path, and cohors remains planned. Adding a provider requires an Auctra OpenSpec change.

Can Auctra export directly to Scaena for video production?▼

Auctra exports only accepted screenplay semantics via the auctra.production_handoff.v1 contract using production handoff commands. It never creates ShotIntent, generation bundles, or provider jobs; Scaena owns execution and Auctra must not mutate .scaena.

Why must run evidence and CLI output avoid prompts and secrets?▼

Redaction is mandatory: secrets, auth headers, raw prompts, provider payloads, and chain-of-thought must not appear in CLI output, receipts, fixtures, golden files, or docs. This keeps local-first .auctra evidence safe to store and share.

Which tests validate Auctra runtime or handoff changes?▼

Run go test -tags=nomsgpack on the affected internal packages plus openspec validate for the relevant change, such as agent-cli-redesign or auctra-shot-production-handoff-v1. Broader changes also require task build, task fixture-demo, and task test:integration.