AGENT SKILL

retail-init

Bootstrap the Compass-Driven kit and lead a NEW user to a first visible result on THEIR own table in the Seshat BI repo. Use when someone asks to "set up the kit", "initialize seshat", "onboard me / get me started", or runs `retail init`. The AGENT performs this: it writes the backstage substrate (compass router + fenced AGENTS.md/CLAUDE.md regions + manifests) via the substrate-only `retail init` CLI, DELEGATES the worked-example offer + human-seam list to `first-hour-compass`, and ROUTES into `retail-onboard-table` for the Stage-1 read-only profile (grain candidates + column types over a live DB, or `[PENDING LIVE PROFILE]` without one). It is NOT a terminal wizard, stores no run-state, fetches no remote, self-grants no approval, and emits no confidence score.

License 许可
Apache-2.0
所属仓库
Kemetra/Seshat-BI
最近更新
2026-11-11
Stars
★ 2
Forks
0

01 前言 Front Matter

name
retail-init
description
Bootstrap the Compass-Driven kit and lead a NEW user to a first visible result on THEIR own table in the Seshat BI repo. Use when someone asks to "set up the kit", "initialize seshat", "onboard me / get me started", or runs `retail init`. The AGENT performs this: it writes the backstage substrate (compass router + fenced AGENTS.md/CLAUDE.md regions + manifests) via the substrate-only `retail init` CLI, DELEGATES the worked-example offer + human-seam list to `first-hour-compass`, and ROUTES into `retail-onboard-table` for the Stage-1 read-only profile (grain candidates + column types over a live DB, or `[PENDING LIVE PROFILE]` without one). It is NOT a terminal wizard, stores no run-state, fetches no remote, self-grants no approval, and emits no confidence score.

03 SKILL.md 正文

retail-init

The install-time front door for the Compass-Driven kit. Its one job is to turn a fresh pip install into a first visible result on the user's own table — leading with analyst value, keeping the substrate backstage. You (the agent) perform this flow; the retail init CLI only writes substrate and prints the next step. This is Phase-1 Step 1-2 of docs/roadmap/distribution-ideas.md.

Scope + non-negotiables (read first)

  • Agent-first, not a wizard (Principle I). The retail init CLI is substrate-writing ONLY — it writes .seshat/ + the fenced regions and prints the next step. It never prompts, shows a menu, or emits a profile. The delegate → route → profile flow is YOU performing this skill over the existing prose verbs.
  • Delegate, never fork (anti-fork). The worked-example offer, the human-seam list, and the single-table orientation card belong to first-hour-compass — its single source. The Stage-1 profile belongs to retail-onboard-table. Do NOT restate any of them here; route into them.
  • No run-state (FR-005). compass.yaml declares the orientation protocol and points at per-table readiness-status.yaml; it stores no current_stage. There is no repo-level stage.
  • Fence-only writes (FR-006/FR-007). The substrate write touches only the <!-- SESHAT-KIT START -->…<!-- SESHAT-KIT END --> region of AGENTS.md / CLAUDE.md; everything outside is hand-authored / constitution-owned and stays byte-identical.
  • Stop at judgment (Principle V). Grain, PII publish-safety, business rollup/segment, product identity are the human's — surfaced (via first-hour-compass) and STOPPED on, never self-granted.
  • ASCII only, UTF-8 no BOM.

The flow (what you perform)

  1. Bootstrap the substrate. Run retail init (or retail init --repo <path>). It writes .seshat/compass.yaml + manifests and projects the SESHAT-KIT fenced regions of AGENTS.md / CLAUDE.md, then prints the next step. Do NOT narrate the substrate as steps — it is backstage.
  2. Set expectations honestly (FR-009). State up front: the agent handles the sequence and the plumbing; you still own the judgment seams. Surface the seam wording from first-hour-compass (its single source) — do not re-type a divergent list.
  3. Delegate the worked-example offer. Route into first-hour-compass, which presents retail-store-sales (full seven-stage spine, docs/worked-examples/ retail-store-sales.md) and takes the user's pick as a narrative pattern to steer by (not a file template).
  4. Route into the profile. Hand the user's named table to retail-onboard-table (the Source → Mapping front door, which owns the Stage-1 read-only profile). The agent-visible first result is the grain candidates + column types it returns.
  5. Degrade honestly when there's no DB. The Stage-1 profile is DB-backed (profile.py over a QueryRunner). If no db extra / DSN is configured, report the boundary and the enable steps (pipx inject seshat-bi psycopg2-binary or pip install "seshat-bi[db]"; set DATABASE_URL or ANALYTICS_DB_* in the gitignored .env), mark profile numbers [PENDING LIVE PROFILE], author the source-map / orientation structure, and STAY USEFUL — never traceback, never fake a pass. There is no CSV/Excel profiler (YAGNI).

What init does NOT do

  • Does not profile, open a DB, prompt, or show a menu (that is the CLI's hard boundary — substrate only).
  • Does not advance a readiness stage, write an approvals[] entry, or emit a numeric health / confidence / percent-ready score (FR-010; hard rule #9).
  • Does not fetch from a remote or auto-execute pulled content — kit self-update (sync, Phase-3) and channel-driven fetch (Phase-4) are later, gated slices (FR-011).
  • Does not rewrite any hand-authored / constitution-owned content; on a malformed fence it STOPS and reports rather than force a rewrite.

Idempotent re-run

Running retail init again on a bootstrapped repo re-projects only the fenced regions, reports "already bootstrapped", and never duplicates a fence or clobbers the hand-authored region.

See also

  • The delegated verbs: .claude/skills/first-hour-compass/SKILL.md (offer + seams + card), .claude/skills/retail-onboard-table/SKILL.md (Stage-1 profile).
  • The router it writes: .seshat/compass.yaml; its canonical source: .seshat/kit-source.yaml.
  • The design + resolved decisions: docs/roadmap/distribution-ideas.md; specs/070-retail-init-bootstrap/.
  • The conductor for the later medallion stages: .claude/skills/retail-orchestrate/SKILL.md.

04 同仓库更多 Skills

Kemetra
Kemetra/Seshat-BI

fake-skill

A fixture skill with declaring frontmatter, used as positive shipped evidence.

★ 2
Kemetra
Kemetra/Seshat-BI

fake-skill

A fixture skill with declaring frontmatter, used as positive shipped evidence.

★ 2
Kemetra
Kemetra/Seshat-BI

bare-skill-no-frontmatter

暂无简介

★ 2
Kemetra
Kemetra/Seshat-BI

fake-skill

A fixture skill with declaring frontmatter, used as positive shipped evidence.

★ 2
Kemetra
Kemetra/Seshat-BI

retail-kpi-knowledge

Retail KPI / metric-contract reasoning layer for BI agents in the Seshat BI project. Use when an agent must reason about the BUSINESS MEANING of a retail KPI before any implementation — defining a KPI in business terms, classifying additivity (fully/semi/non-additive), identifying grain, listing required source fields, resolving ambiguity (gross vs net, VAT, returns, cost method, same-store), choosing an MVP KPI pack, reviewing a metric contract for completeness, or preparing a clean handoff to an implementation layer. Also use it to classify KPI sufficiency or prepare an owner policy decision packet. This is a meaning/definition + review layer, not an executor: it does not write DAX/SQL/Python, grant readiness, approve policy, or design dashboards. See INDEX.md for live/planned routes; registry.yaml is the authoritative machine-readable inventory.

★ 2
Kemetra
Kemetra/Seshat-BI

bi-sql-knowledge

SQL reasoning, validation, reconciliation, transformation, diagnostics, and anti-pattern knowledge layer for Seshat BI agents. Use when a task involves reasoning about source SQL: source profiling, table grain, keys & uniqueness, joins & fan-out amplification, aggregation correctness, COUNT/NULL semantics, deduplication, validation queries, reconciliation queries, silver/gold transformation logic, window/date-time analytics, SQL anti-patterns, or basic SQL performance reasoning, or interpretation of a supplied sanitized PostgreSQL JSON execution plan. The SQL counterpart to bi-dax-knowledge. It is a thinking & validation layer, not a database executor, runtime validator, dbt/Dagster project, or automatic tuning service.

★ 2