agentleFS
ray5273ray5273/kr-research-kit/AGENTS.md

kr-research-kit

AGENTS.md2 starsChanged 8 days agoMITView on GitHub

The file

We split this AGENTS.md into the rules it gives an agent

12 rules. About 0 are common, 6 are distinctive, and 6 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈6

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 6

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    When you add or update a stock analysis example, keep it under analysis-example/<market>/<company>/.

    line 10
  2. Very specific:

    When a stock-analysis skill is invoked in this repo and the workspace is writable, create or update the matching report file under …

    line 11
  3. Distinctive:

    When a commit changes a skill's required output shape, refresh any affected example documents so they still match the current skill …

    line 13
  4. Distinctive:

    When an example document embeds a chart image, refresh the linked PNG asset together with the markdown file so the visual stays in sync …

    line 14
  5. Very specific:

    When you update an existing Korean memo with follow-up analysis, preserve the original 기준일, refresh 최근 업데이트일, and append or replace a dated …

    line 15
  6. Distinctive:

    Treat example analysis documents as time-sensitive.

    line 17
  7. Distinctive:

    If the example must mix metrics from sources with different refresh cadences, keep the document honest by stating the exact date for each …

    line 18
  8. Distinctive:

    Do not invent revenue mix, customer concentration, governance facts, or valuation metrics that are not disclosed in the current source set.

    line 19
  9. Very specific:

    Do not invent foreign-IB views.

    line 20
  10. Distinctive:

    Prefer primary sources first for Korean stocks: DART, KRX disclosures, company IR pages, governance pages, and official financial …

    line 23
  11. Very specific:

    Before publishing a new or edited article, use skills/kr-article-readability/SKILL.md to review emphasis, tone colors, paragraph length, …

    line 28
  12. Very specific:

    Run npm run check:readability before site or preview deployment.

    line 29

≈ “Common” is an estimate: at least 1 in 5 of the 40,246 AGENTS.md files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
AGENTS.md

AGENTS

External Agent Targets

  • A request referring to the Hermes agent means the remote Hermes runtime (SSH address in ~/stock-analysis-private/HERMES.md, not in this public repo) unless the user explicitly names another target.
  • Inspect the remote runtime and its active configuration before changing Hermes behavior. Keep source and the installed ~/.hermes runtime copy aligned when a scheduled Hermes script is changed.

Repository Rules

  • When you add or update a stock analysis example, keep it under analysis-example/<market>/<company>/.
  • When a stock-analysis skill is invoked in this repo and the workspace is writable, create or update the matching report file under analysis-example/<market>/<company>/memo.md instead of leaving the analysis only in chat.
  • When you add a new example file, do NOT append a link to README.md. Stage the example first (git add), then run node scripts/build-artifact-index.js to regenerate docs/ARTIFACTS.md, and commit the index together with the example. The index deliberately covers only git-tracked files so its links never 404 on a fresh clone. The README carries only a handful of curated representative links; replace one only when the new artifact is a better canonical demonstration, never by appending.
  • When a commit changes a skill's required output shape, refresh any affected example documents so they still match the current skill behavior.
  • When an example document embeds a chart image, refresh the linked PNG asset together with the markdown file so the visual stays in sync with the text.
  • When you update an existing Korean memo with follow-up analysis, preserve the original 기준일, refresh 최근 업데이트일, and append or replace a dated block under ## Update Log instead of rewriting the whole memo by default.
  • For dated follow-up updates, search only for company-specific disclosures, IR materials, and news published after the memo's 기준일, then deduplicate against existing update dates and source URLs when possible.
  • Treat example analysis documents as time-sensitive. Before committing changes to an example document, refresh prices, valuation metrics, disclosures, and source dates with current sources.
  • If the example must mix metrics from sources with different refresh cadences, keep the document honest by stating the exact date for each metric block or row.
  • Do not invent revenue mix, customer concentration, governance facts, or valuation metrics that are not disclosed in the current source set. State what is not separately disclosed.
  • Do not invent foreign-IB views. Every ## Street / Alternative Views bullet that names a foreign broker, rating, or target price must trace back to a dated Korean news article that quoted that broker; prefer kr-foreign-analyst output over paraphrasing from memory.
  • When the user asks for a Korean post-earnings review (실적 리뷰, 실적발표 자료), read and follow skills/kr-earnings-review/SKILL.md end to end: run collect-earnings-inputs.js, pass its readiness gate, and if any required input is missing stop in 검토 대기 with .tmp/earnings/<ticker>-<period>/REVIEW-NEEDED.md for the user's review instead of publishing. At the start of a session, check .tmp/earnings/*/REVIEW-NEEDED.md for reviews left on hold.
  • For a U.S.-listed company's post-earnings review, follow skills/us-earnings-review/SKILL.md the same way (collect-us-earnings-inputs.js, the same R1–R6 gate, .tmp/earnings/us-<TICKER>-<period>/REVIEW-NEEDED.md when on hold).
  • Prefer primary sources first for Korean stocks: DART, KRX disclosures, company IR pages, governance pages, and official financial statements.
  • When the memo needs sell-side consensus (target price, rating distribution, broker coverage), use the analyst-report chain (kr-analyst-report-discover → kr-analyst-report-fetch → kr-analyst-report-insight). It pulls reports from consensus.hankyung.com (primary) with finance.naver.com/research as fallback, defaults to a 365-day lookback, and skips login-gated reports. Treat the resulting analyst-report-insight.md digest as sell-side inference, not a primary filing fact — it lands in analysis-example/kr/<company>/ and is ingested into kr-stock-data-pack's External Views with Source role: sell-side consensus.

Article readability before site deployment

  • Before publishing a new or edited article, use skills/kr-article-readability/SKILL.md to review emphasis, tone colors, paragraph length, section rhythm, tables and source labels. Preserve facts, source dates and judgment in layout-only edits.
  • Run npm run check:readability before site or preview deployment. Official npm deployment commands also run this gate automatically. A new/changed article needs a matching review in site/readability-reviews.json; do not generate review records without doing the editorial review.
  • Legacy article snapshots are explicitly unreviewed debt, not passes. Do not bulk-refresh those hashes to bypass the review. If browser checks are unavailable, record source/renderer checks honestly and retain the visual limitation.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.