How do I install the Extract Advisorhub Articles skill?

Run npx skillmds@latest add codyswanngt/extract-advisorhub-articles in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

Is Extract Advisorhub Articles safe to use?

This skill has not completed SkillMD's automated safety review yet. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with Extract Advisorhub Articles?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is Extract Advisorhub Articles free to use?

Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.

Who published Extract Advisorhub Articles?

codyswanngt (@codyswanngt) published this skill. Their other Agent Skills are listed on their SkillMD profile.

Extract Advisorhub Articles by codyswanngt

Name: Extract Advisorhub Articles
Author: codyswanngt

Extract Advisorhub Articles

Extract structured advisor data from saved AdvisorHub wp-json articles and load it into Harper using the repo's Bun/TypeScript helper and loader scripts. Use for richer entities than regex ingest: advisors, firms, teams, transitions, disclosures, sanctions, employment histories, OBAs, and field assertions. Safe to re-run because pending detection and loader IDs are deterministic.

codyswanngt 6ef8ce7 3 files · 37.4 KB Updated Sep 17, 2026

File contents

Extract AdvisorHub Articles -> Harper

This is the rich-extraction phase that runs after ingest-advisorhub has populated research/wpjson/.

Pipeline:

research/wpjson/**/post_*.json
  -> Codex reads article prose and writes research/extractions/<wpId>.json
  -> bun run load:extractions
  -> Harper upserts Article, Advisor, Firm, Team, Disclosure, etc.
  -> loaded files move to research/extractions/.loaded/

Extraction JSON files are runtime artifacts and are gitignored.

Pre-flight

Use Bun:

bun run build
bun run status

If Harper is stopped, ask before bun run bootstrap. If research/wpjson/ is empty, run ingest-advisorhub first.

For deployed Harper, set HDB_TARGET_URL, HDB_ADMIN_USERNAME, and HDB_ADMIN_PASSWORD; otherwise the loader uses the local Harper socket.

Find Pending Articles

bun run build
node dist/scripts/extract_helper.js find-pending

The output is tab-separated:

<wpId> <source-file> research/extractions/<wpId>.json

For a timer/Automation, process a capped batch. A good default is 10 articles per run; use 20 only when articles are short and context is comfortable.

Extract Each Article

Show an article:

node dist/scripts/extract_helper.js show <wpId>

Before writing the first extraction in a batch, read:

.agents/skills/extract-advisorhub-articles/schema-guide.md
.agents/skills/extract-advisorhub-articles/examples.md

Write the result to:

research/extractions/<wpId>.json

Rules:

Every entity must have a natural_key.
Every factual field must have a matching field_assertions entry.
Every assertion quote must appear in the article text.
Use confidence: "asserted" for verbatim facts, "inferred" for facts derived from prose, and "derived" for computed values.
Cross-references use natural-key fields such as advisor_legal_name, firm_canonical_name, team_name, and disclosure_local_key; do not invent UUIDs in extraction JSON.
If an article has no extractable people/firms/events, mark article.has_extractable_content: false and keep entity arrays empty.

Load

Load all pending extraction files:

bun run load:extractions

Load one article:

bun run load:extractions -- --wpid <wpId>

Dry-run:

bun run load:extractions -- --dry-run

The loader resolves deterministic IDs via src/lib/ids.ts, upserts through src/lib/harper.ts, and moves successfully loaded files to research/extractions/.loaded/.

Confirm

bun run verify

or, against deployed Harper:

bun run verify:rest

Report:

number of pending articles found and extracted
loader summary per table
articles skipped as non-extractable
validation issues, ambiguous identity matches, or Harper access failures

Automation Guidance

For a recurring Codex Automation, use a prompt that:

builds first with bun run build
runs node dist/scripts/extract_helper.js find-pending
processes at most 10 pending articles
loads with bun run load:extractions
verifies with bun run verify or bun run verify:rest
reports concise table counts and skipped articles

Do not schedule this as an unlimited loop; rich extraction consumes agent context and should advance in small batches.

---name: extract-advisorhub-articlesdescription: Extract structured advisor data from saved AdvisorHub wp-json articles and load it into Harper using the repo's Bun/TypeScript helper and loader scripts. Use for richer entities than regex ingest: advisors, firms, teams, transitions, disclosures, sanctions, employment histories, OBAs, and field assertions. Safe to re-run because pending detection and loader IDs are deterministic.---# Extract AdvisorHub Articles -> HarperThis is the rich-extraction phase that runs after`ingest-advisorhub` has populated `research/wpjson/`.Pipeline:```textresearch/wpjson/**/post_*.json -> Codex reads article prose and writes research/extractions/<wpId>.json -> bun run load:extractions -> Harper upserts Article, Advisor, Firm, Team, Disclosure, etc. -> loaded files move to research/extractions/.loaded/```Extraction JSON files are runtime artifacts and are gitignored.## Pre-flightUse Bun:```bashbun run buildbun run status```If Harper is stopped, ask before `bun run bootstrap`. If`research/wpjson/` is empty, run `ingest-advisorhub` first.For deployed Harper, set `HDB_TARGET_URL`, `HDB_ADMIN_USERNAME`, and`HDB_ADMIN_PASSWORD`; otherwise the loader uses the local Harper socket.## Find Pending Articles```bashbun run buildnode dist/scripts/extract_helper.js find-pending```The output is tab-separated:```text<wpId> <source-file> research/extractions/<wpId>.json```For a timer/Automation, process a capped batch. A good default is 10articles per run; use 20 only when articles are short and context iscomfortable.## Extract Each ArticleShow an article:```bashnode dist/scripts/extract_helper.js show <wpId>```Before writing the first extraction in a batch, read:- `.agents/skills/extract-advisorhub-articles/schema-guide.md`- `.agents/skills/extract-advisorhub-articles/examples.md`Write the result to:```textresearch/extractions/<wpId>.json```Rules:- Every entity must have a `natural_key`.- Every factual field must have a matching `field_assertions` entry.- Every assertion quote must appear in the article text.- Use `confidence: "asserted"` for verbatim facts, `"inferred"` for facts derived from prose, and `"derived"` for computed values.- Cross-references use natural-key fields such as `advisor_legal_name`, `firm_canonical_name`, `team_name`, and `disclosure_local_key`; do not invent UUIDs in extraction JSON.- If an article has no extractable people/firms/events, mark `article.has_extractable_content: false` and keep entity arrays empty.## LoadLoad all pending extraction files:```bashbun run load:extractions```Load one article:```bashbun run load:extractions -- --wpid <wpId>```Dry-run:```bashbun run load:extractions -- --dry-run```The loader resolves deterministic IDs via `src/lib/ids.ts`, upsertsthrough `src/lib/harper.ts`, and moves successfully loaded files to`research/extractions/.loaded/`.## Confirm```bashbun run verify```or, against deployed Harper:```bashbun run verify:rest```Report:- number of pending articles found and extracted- loader summary per table- articles skipped as non-extractable- validation issues, ambiguous identity matches, or Harper access failures## Automation GuidanceFor a recurring Codex Automation, use a prompt that:- builds first with `bun run build`- runs `node dist/scripts/extract_helper.js find-pending`- processes at most 10 pending articles- loads with `bun run load:extractions`- verifies with `bun run verify` or `bun run verify:rest`- reports concise table counts and skipped articlesDo not schedule this as an unlimited loop; rich extraction consumesagent context and should advance in small batches.

Frequently asked questions