# Ingest

> |

- **Type:** Skill
- **Install:** `agentstack add skill-kouko-monkey-skills-ingest`
- **Verified:** Yes — security-reviewed for prompt injection and unsafe behavior
- **Seller:** [kouko](https://agentstack.voostack.com/s/kouko)
- **Installs:** 0
- **Category:** [Agent Skills](https://agentstack.voostack.com/c/agent-skills)
- **Latest version:** 0.1.0
- **License:** MIT
- **Upstream author:** [kouko](https://github.com/kouko)
- **Source:** https://github.com/kouko/monkey-skills/tree/main/dbt-wiki/skills/ingest

## Install

```sh
agentstack add skill-kouko-monkey-skills-ingest
```

Requires the [AgentStack CLI](https://agentstack.voostack.com/docs/cli). Works with Claude Code, Cursor, and any MCP-compatible agent.

## About

# dbt-wiki — Ingest Workflow (v1.0)

User-supplied context capture. Mirrors `/repo-wiki:ingest` context
mode for the dbt-internal scope. Notes attach to specific
model / source / macro pages and survive `/dbt-wiki:rescan` runs
(preserved as user-owned `## User Notes` body section per SCHEMA.md).

dbt-wiki ingest is **context-only** in v1.0. It does NOT have:
- `git` mode (manifest.json snapshot supersedes git for dbt structure;
  use /dbt-wiki:rescan after `dbt parse` instead)
- `doc-import` mode (v2 backlog — for now, paste the doc summary as
  context manually, or write to a `docs/` file and reference it)

## Pre-condition Check

```bash
test -d .dbt-wiki || { echo "Knowledge base not initialized. Run /dbt-wiki:init first."; exit 1; }
test -f .dbt-wiki/index.md || { echo "Missing .dbt-wiki/index.md — re-run /dbt-wiki:init."; exit 1; }
```

If user invocation has no argument:

> Usage: `/dbt-wiki:ingest ""`
>
> Example: `/dbt-wiki:ingest "fct_orders sort_key is (order_date, customer_id) because Tableau extract joins on these — see incident #4521"`
>
> The note will be attached to the relevant model / source / macro
> page based on names mentioned in your text. If multiple matches
> or no match is detected, you'll be asked to clarify.

Exit cleanly if no arg.

## Step 1: Identify Target Resource(s)

Parse the user's argument text for resource names. Match against
existing pages under:
- `.dbt-wiki/_evidence/models/*.md` — model names
- `.dbt-wiki/_evidence/sources/*.md` — source names (in form `__`)
- `.dbt-wiki/_evidence/macros/*.md` — macro names
- `.dbt-wiki/_evidence/seeds/*.md`, `_evidence/snapshots/*.md`, `_evidence/exposures/*.md`

Matching strategy:
1. **Exact name match**: tokens in the argument that exactly match a
   resource name (case-sensitive, since dbt is case-sensitive)
2. **Tier-prefix match**: if user mentions a tier ("marts_finance 整層"),
   collect all models under that path prefix
3. **No match → ask**:

   > Couldn't identify a target resource in your note. Available models:
   > 
   >
   > Which model / source / macro should this note attach to?

4. **Multiple matches** (when user mentions 2+ resources, or tier-prefix
   matches multiple):

   > Your note mentions . Attach to:
   > [a] all of them (same note copied to each)
   > [b] just  — pick one
   > [c] cancel

## Step 2: Compose Note Entry

For each target resource:

```markdown
###  

```

Slug rule: take first 6-8 words of the user's text, kebab-case, lowercase.
Skip dbt resource names already in heading. Examples:

| User text | Slug |
|---|---|
| "fct_orders sort_key is (order_date, customer_id) because..." | `sort-key-rationale` |
| "marts_finance permissions need analytics_readonly_group..." | `permissions-readonly-group` |
| "stg_customers has GDPR redaction merge — see incident #4521" | `gdpr-redaction-merge` |

If slug collides with an existing entry on the same date, append `-2`, `-3`, ...

## Step 3: Append to Target Page

Read the target page. Locate the `## User Notes` section:
- If exists: append the new entry below the last existing entry
- If missing: insert `## User Notes` after `## Tests` section (per SCHEMA.md
  standard ordering); then add the entry

NEVER:
- Modify any other section (Description / SQL Preview / Column Sources /
  Tests / Inline Comments / Cross-references / etc.)
- Remove existing User Notes entries
- Re-order existing entries
- Change the section heading text

## Step 4: Append Log Entry

```
## [] ingest | attached note to  page(s)
- Targets: [, ...]
- Slug: 
- Note size:  chars
```

## Step 5: Summary

```
✓ Note attached to  page(s):
  - .dbt-wiki/_evidence/models/fct_orders.md  (### 2026-05-02 sort-key-rationale)

  To verify: /dbt-wiki:query "fct_orders sort key 為什麼這樣設？"
  To add more: /dbt-wiki:ingest ""
```

## Cross-skill Integration

If `.repo-wiki/` exists in the same repo, suggest the user consider
whether the note belongs in dbt-wiki (dbt-internal: model behavior,
column quirks, materialization) or in repo-wiki (cross-cutting WHY:
business decisions, project-level history):

> Note attached to dbt-wiki. If this context is relevant beyond dbt
> (e.g., business-level decision touching multiple subsystems), consider
> also: `/repo-wiki:ingest ""` to capture at repo level.

This is informational only — do NOT auto-cross-post.

## Rules

NEVER:
- Modify any file outside `.dbt-wiki/.md` (specifically: never
  touch `dbt/`, `target/`, or other plugins' directories)
- Rewrite or paraphrase the user's note content (light whitespace
  normalization OK; semantic editing NOT OK)
- Auto-attach without confirming target on multi-match or no-match
- Skip the `## User Notes` section heading (rescan recognizes it as
  user-owned by name)
- Use `[[wikilinks]]` — only standard markdown links
- Connect to dbt Cloud / warehouse / external API

ALWAYS:
- Verify `.dbt-wiki/` exists before any work
- Match resource names case-sensitively (dbt convention)
- Append, never replace, existing User Notes entries
- Date-prefix every new entry (`### YYYY-MM-DD `)
- Preserve user's original wording and language (zh / ja / en mixed OK)
- Append a log entry per ingest invocation
- Suggest /repo-wiki:ingest cross-link when note has cross-cutting scope
  (informational; never auto-post)

## Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

- **Author:** [kouko](https://github.com/kouko)
- **Source:** [kouko/monkey-skills](https://github.com/kouko/monkey-skills)
- **License:** MIT

Install and usage instructions live in the source repository linked above.

## Pricing

- **Free** — Free

## Security capabilities

Automated source analysis of v0.1.0 — what this tool can access:

- **Network access:** no
- **Filesystem access:** no
- **Shell / process execution:** no
- **Environment & secrets:** no
- **Dynamic code execution:** no

*"Yes" means the capability is present in the source — more access means more to trust, not that it is unsafe.*


## Versions

- **0.1.0** — security scan: passed — Imported from the upstream source.

## Links

- Listing page: https://agentstack.voostack.com/l/skill-kouko-monkey-skills-ingest
- Seller: https://agentstack.voostack.com/s/kouko
- Browse the marketplace: https://agentstack.voostack.com/browse

---
Listed on AgentStack — the marketplace for AI agent skills and MCP servers. Every listing is security-reviewed. Creators keep 70%.