nave-ism-ranking-workflow

name	nave-ism-ranking-workflow
description	Generate ISM manufacturing/services rankings using official /root/nave commands and snapshots.
version	1.0.0
author	Hermes Agent
license	MIT
metadata	{"hermes":{"tags":["nave","ism","ranking","trading","cli"]}}

nave ISM ranking workflow

When to use

Use this workflow whenever the user asks for ISM rankings (current month or prior months) for manufacturing and/or services in the nave project.

Rule

Always use official repo commands from /root/nave. Do not use ad-hoc web-only ranking methods as the primary source.

Commands (official)

Run from repo root:

cd /root/nave
/root/nave/.venv/bin/nave stocks ism-report --kind manufacturing --mode manufacturing --top-n 15 --json
/root/nave/.venv/bin/nave stocks ism-report --kind services --mode services --top-n 15 --json

These commands produce:

• ISM report month + PMI
• hottest/worst industries
• long/short candidates with score/confidence
• saved_to path for monthly snapshot JSON

Snapshot locations

By default snapshots are written to:

• stocks_history/ism_manufacturing_YYYY-MM.json
• stocks_history/ism_services_YYYY-MM.json

For prior-month review, load those snapshot files first if they exist.

Suggested readback checks

1. Verify report_month in JSON output.
2. Verify saved_to path exists.
3. Quote recommendations from candidates.longs / candidates.shorts and include confidence, score, and reason.
4. For any user-requested "último ISM" report (manufacturing or services), also run the official X/Twitter analysis for the candidate tickers from saved_to before answering:

   cd /root/nave
   set -a; . /root/nave/.env; set +a
   NAVE_CLI_STATUS=0 /root/nave/.venv/bin/nave stocks x-analyze --from-snapshot <saved_to> --top 5 --days 7 --limit-per-ticker 50 --json
   

   Synthesize the X payload per ticker: useful points of view, sentiment/crowding, catalysts/risks being discussed, and whether the chatter supports waiting for pullback/trigger vs entry now. Do not replace the official ISM ranking with X sentiment; X is a secondary due-diligence layer.
5. If the user asks "cuál comprar" from an ISM report, add an execution filter instead of just repeating the ranking:
   • Exclude/adjust tickers the user already bought or says they already own.
   • Check live Ondo availability if the user is likely trading tokenized stocks: fetch https://app.ondo.finance/api/v2/assets with User-Agent: Mozilla/5.0, map ticker, underlyingMarket.ticker, and symbol minus the on suffix, then separate "best overall" from "best if only Ondo".
   • Add a basic technical overlay for shortlisted tickers using Yahoo chart data: last price, SMA20/SMA50/SMA200, distance from 52-week high/low, and whether the ticker is extended.
   • Prefer a concrete buy/entry zone, invalidation, and initial target over a generic ranking. Flag names near 52-week highs as potentially extended before recommending them.
6. Flag names near 52-week highs as potentially extended before calling them "rezagadas".
7. When the user asks for a month around a release boundary, distinguish covered month from publication date. Example: Manufacturing April 2026 was published on May 1, 2026, but snapshot/report month is 2026-04 / April 2026; Manufacturing May 2026 is published in June.

Calendar verification

Use the official repo calendar commands when month availability or release timing is ambiguous:

cd /root/nave
/root/nave/.venv/bin/nave stocks ism-calendar show --json
/root/nave/.venv/bin/nave stocks ism-calendar next --json

Check covers_month, release_date, and event. If the requested covered month has not been released yet, say so explicitly and fall back to the latest valid snapshot/report only if useful.

Pitfalls

• ISM naming is easy to confuse: "the May 1 release" usually covers April; "May report" may mean either publication month or covered month. Confirm using ism-calendar before correcting the user or labeling a report unavailable.
• nave stocks ism-report output can be large; terminal may truncate. Re-run with output redirected to file when needed.
• Some candidates appear on both expanding/contracting sector universes due sector overlap; rely on final candidate side in output.
• Data quality depends on upstream provider fields (yfinance/fmp), so keep source fields in explanations.

Release-day automation with Hermes cron

When automating ISM release-day reports for the user, schedule a daily afternoon Argentina-time cron that first checks the official Nave calendar and exits quietly/minimally when there is no release that day. Do not hardcode release dates as the primary mechanism.

Canonical automated sequence:

1. Determine today's date in America/Argentina/Buenos_Aires (do not assume the VPS local timezone is Argentina).
2. From /root/nave, load repo credentials before running stock workflows:

   set -a; . /root/nave/.env; set +a
   

3. Run NAVE_CLI_STATUS=0 /root/nave/.venv/bin/nave stocks ism-calendar show --json and parse from the first { because the CLI may print a [data_loader] status line before JSON.
4. If release_date == today, run the matching official report command:

   NAVE_CLI_STATUS=0 /root/nave/.venv/bin/nave stocks ism-report --kind manufacturing --mode manufacturing --top-n 15 --json
   NAVE_CLI_STATUS=0 /root/nave/.venv/bin/nave stocks ism-report --kind services --mode services --top-n 15 --json
   

5. Verify report_month / saved_to matches the calendar covers_month; if not, report that the new ISM is not available yet and retry later rather than using an old month as if it were new.
6. Run the official X/Twitter add-on from the generated snapshot:

   NAVE_CLI_STATUS=0 /root/nave/.venv/bin/nave stocks x-analyze --from-snapshot <saved_to> --top 5 --days 7 --limit-per-ticker 50 --json
   

7. Summarize in Spanish: covered month, PMI if available, strong/weak industries, long/short candidates with score/confidence/reason, X sentiment by ticker, entry/trigger/invalidation notes when present.

See references/hermes-cron-ism-release-day.md for a known-good cron prompt pattern and operational notes.

Verification

• Command exits 0.
• Output includes keys: report_month, hottest_industries, worst_industries, candidates, saved_to.
• Snapshot file exists at saved_to.
• On release-day automation, calendar covers_month matches report report_month before presenting the ranking as the new report.