Trump Truthsocial Catalyst

Trump Truth Social Catalyst Lens

A reusable lens for turning Donald Trump's public Truth Social posts, @realDonaldTrump X posts, and his own disclosed stock holdings into a short list of US stocks/sectors likely to move, with an honest read on how reliably each pattern actually pumps a stock. Built to be paired with a periodic timer that re-fetches and re-scores, accumulating a real hit-rate over time.

Decision-support lens, NOT financial advice and NOT an auto-trader. This

only analyzes publicly available posts and public financial disclosures to

anticipate market reactions. It never places, cancels, or sizes orders. Always

confirm current price/news yourself — these moves are fast, often fade, and

several are pump-and-fade traps (see references/track-record.md).

---

STEP 0 — ALWAYS REFRESH FIRST (this skill is useless on stale data)

Two things go stale: the skill itself (the ledger / holdings / calibration in references/ are grown by a periodic timer and pushed to GitHub) and the posts (Trump posts many times a day; a market-moving one can be minutes old, and there is no local post archive). Refresh both, in order, before doing anything else:

# 0a. Pull the latest version of this skill (updated ledger, holdings, calibration).
#     Best-effort: auto-detects project vs global scope. If it fails, keep going.
skills update trump-truthsocial-catalyst -y

# 0b. Fetch the latest Truth Social posts: text + timestamps + permalinks.
#     trumpstruth.org is a third-party archive; the official truthsocial.com API is
#     Cloudflare-blocked from servers, so use this RSS.
curl -sS -A "Mozilla/5.0" "https://trumpstruth.org/feed" -o /tmp/trump_feed.xml

# 0c. Fetch the latest @realDonaldTrump X posts when xreach is available.
#     X is a supplemental official public channel; many posts are video/link-only,
#     so resolve context before scoring.
xreach tweets @realdonaldtrump -n 40 --json > /tmp/trump_x.json

Then *re-read the refreshed `references/.md** (don't rely on a copy you read before 0a) and parse the newest posts (see references/methodology.md` for ready parsers). If 0a fails, proceed on the cached skill but note it may be stale; if a source fails (offline / 403 / site down), say so explicitly and use the other source plus whatever the user pasted — never pretend you have fresh data.

---

What this lens is for

Trump's posts move stocks through a few repeatable mechanisms: he names a company favorably, announces or hints at policy he controls (tariffs, contracts, executive orders), or champions a theme (crypto, energy, AI infrastructure, defense, reshoring). The edge is recognizing which features of a post reliably pump a stock versus which produce a one-day spike that fades — and cross-referencing against what he personally owns, because "he owns it → he boosts it → a government tailwind follows" is his single most repeatable money-making pattern (the Dell playbook).

This is explicitly a short-horizon catalyst lens, not a fundamentals lens. The biggest caveat in the whole skill: his posts can pump a stock that is a terrible long-term hold — his own DJT is down ~77% from its 2024 high and his memecoins are down 97–99%, yet they still spike when he posts. Separate "will it pop this week" from "is it a good investment."

---

How the reference files are organized

Read progressively — pull in only what the task needs.

---

Workflows

(a) Read one post (or a batch) for tradable signals

1. Refresh (Step 0). Identify the newest Truth Social and X posts; dedupe against anything already in references/track-record.md by source id and cross-source normalized text/time. 2. For each post, run the classification checklist in references/methodology.md: is it market-relevant? Which Tier (1 named-stock / 2 policy-sector / 3 macro)? Which amplifier features are present (named public company, explicit "BUY", policy he controls, he personally owns it, posted intraday, repeated)? 3. Map topic → candidate tickers via references/theme-ticker-map.md, and check references/holdings-watchlist.md for an ownership overlap (raises conviction). 4. Output, per signal: the post (date + permalink), the implicated ticker(s), direction, a reliability tier, expected magnitude & durability (intraday-fade vs. can-hold), and what would invalidate it. 5. Weight using references/track-record.md — favor patterns with a real accumulated hit-rate over plausible-sounding one-offs. Always frame as analysis, never as an order.

(b) Build / refresh a Trump-catalyst watchlist

1. Pull his disclosed holdings from references/holdings-watchlist.md. If the question depends on freshness, check the OGE source listed there for newer Trump, Donald J 278-T PDFs before treating the watchlist as current. 2. Bucket: already popped (Dell-style, catalyst spent) vs. owns-but-quiet (breakout candidates if the Dell template repeats: ownership + an admin tailwind like a contract, tariff carve-out, or public praise). 3. For each quiet name, note the specific catalyst that would trigger it and a rough 1–2 month watch window. Keep it advisory.

(c) Judge a single "will this pump X?" question

1. Refresh, find the relevant post(s). 2. Score the pattern against references/methodology.md tiers + amplifiers and the historical analogs in references/track-record.md. 3. Give a calibrated answer: likely direction, reliability, magnitude, durability, and the honest base-rate (e.g., "named-company endorsements pop reliably intraday; crypto-reserve posts spike then fade next day").

---

Risk & disclaimer framing (state this when giving any view)

• Short-horizon catalyst, not a thesis. A pop is not a reason to hold. His

own assets (DJT, $TRUMP, $MELANIA) show posts can pump structurally weak names.

• Many signals fade fast. Crypto-reserve and tariff-relief posts often spike

for a day and give it back; size and timing matter, and this skill does not set either — the user does.

• Conflict-of-interest is a feature of the data, not an endorsement. He has

bought stocks (Dell, Palantir, Nvidia, media names) that his administration then affected. We note the overlap as a signal, while flagging the obvious manipulation/legal scrutiny around it. Do not act illegally on it.

• Disclosures are ranges and self-reported. Holding figures come from OGE

periodic transaction reports (dollar ranges, not exact amounts) and may lag.

• This is a lens, not a signal feed. It surfaces better questions about a

Trump headline. It is NOT auto-trading, NOT a recommendation, NOT financial advice. Every order is the reader's own manual, confirmed decision.

npx skills add yan-labs/trump-truthsocial-catalyst

<p align="center"><img src="assets/trump.jpg" alt="Donald Trump" width="140" height="140"></p>

trump-truthsocial-catalyst


A decision-support lens that reads Donald Trump's latest Truth Social posts, @realDonaldTrump X posts, and his disclosed personal/family stock holdings as short-term US-stock catalysts: which names or sectors are likely to pop, and an honest read on how reliably each pattern actually does.

It live-fetches the newest posts on every use from the public trumpstruth.org RSS feed and, when available, @realDonaldTrump on X via xreach. The official truthsocial.com API is Cloudflare-blocked from servers, so the RSS archive remains the main high- frequency text source; X is a supplemental official public channel that can surface videos, reposts, and high-engagement mirrors. There is no local post archive — the skill refreshes fresh each time so it never reasons on stale data. It is built to be paired with a periodic timer that re-fetches, re-scores, and accumulates a real, measured hit-rate per pattern over time.

The edge it cross-references: he owns a stock → he praises it → an administration tailwind (a contract, a tariff carve-out) follows → it pops — his single most repeatable pattern (the Dell template).

⚠️ Not financial advice. Decision-support only. This skill never trades and

never places, cancels, or sizes orders. A Trump-driven pop is a *short-horizon

trade setup, NOT a reason to hold*: his own DJT is down ~77% from its 2024 high

and his memecoins ($TRUMP / $MELANIA) are down 97–99% from peak, yet they still

spike when he posts. Conflict-of-interest overlaps (stocks he owns that his

administration then affects) are surfaced as a signal, with the obvious

manipulation/legal scrutiny flagged — never as an endorsement to act illegally.

Holdings come from self-reported OGE disclosures (dollar ranges, may lag).

Always confirm current price/news yourself.

What's in here

How it works

Step 0 — always refresh first. Trump posts many times a day and a market-moving one can be minutes old, so the skill fetches live before doing anything else:

curl -sS -A "Mozilla/5.0" "https://trumpstruth.org/feed" -o /tmp/trump_feed.xml

It parses the newest items (clean text + timestamps + permalinks), dedupes by status_id, and runs each market-relevant post through a classification checklist. Posts are sorted into a tiered pump playbook, ranked by reliability of an immediate up-move and annotated for durability:

• Tier 1 — names a specific company (most reliable). Direct favorable mention

or an explicit all-caps "BUY". Pops immediately and can hold when a real catalyst (a government contract, earnings) follows.

• Tier 2 — policy / crypto / tariff announcements (strong pop, mixed

durability). Mega-investment headlines, tariff exemptions/pauses, pro-crypto reserve posts — reliable direction for a fast trade, but often fade the next day.

• Tier 3 — macro themes (directional, not a single name). "Drill baby drill,"

Fed-cut pressure, defense, reshoring — these shift a whole basket, weaker and slower.

Amplifier features stack the odds — a named public company, a superlative + call to action, a policy he controls, personal ownership (the Dell template), intraday timing, and repetition. Each call is weighted against the historical base-rate in the track-record ledger, then framed as analysis — never as an order.

Install

One-command install with skills.sh:

npx skills add yan-labs/trump-truthsocial-catalyst

Because the skill is most useful on fresh data — and is designed to grow a real hit-rate via a periodic timer — refresh it regularly:

skills update trump-truthsocial-catalyst -y

---

This repository contains only public information: Trump's public Truth Social posts and self-reported public financial disclosures, plus derived analysis. It is an independent research artifact and is not affiliated with, endorsed by, or connected to Donald Trump, Truth Social, or trumpstruth.org. The portrait is the official 2025 White House presidential portrait, a U.S. federal government work in the public domain.

Trump holdings & breakout watchlist

What Trump (his trust) and his family disclosed owning, used as a catalyst watchlist. The key pattern this enables: he owns it → he praises it / a favorable policy follows → it pops (the Dell template). Names he owns that have not yet had a dedicated catalyst are the "not-yet-popped" candidates.

Source & caveats. Figures are from OGE periodic transaction reports and

press coverage (2026). Disclosures report dollar ranges, not exact amounts,

often don't say stock vs. bond, and can lag weeks. "Owns" = disclosed a

transaction; he may have since sold. Self-reported, unverified. This is a

conflict-of-interest map used as a signal, not an endorsement of acting on it.

Authoritative OGE disclosure source

Use the U.S. Office of Government Ethics (OGE) Officials' Individual Disclosures Search Collection as the primary source for new public Trump financial-disclosure documents:

• Search page:

https://www.oge.gov/web/oge.nsf/Officials%20Individual%20Disclosures%20Search%20Collection?OpenForm

• Backing DataTables endpoint:

https://extapps2.oge.gov/201/Presiden.nsf/API.xsp/v2/rest

• Query pattern: send the same DataTables parameters used by the page, filter

columns[3][search][value]=Trump, sort docDate descending, and look for rows where name is Trump, Donald J and type includes direct PDF links such as 278 Transaction.

• Implementation note: include the six page columns (docDate, title, type,

name, agency, level) and each column's name, searchable, orderable, search[value], and search[regex] parameters. A bare endpoint call or global search[value]=Trump is not enough; the useful filter is the name column filter.

• Example verified latest 2026 result set: the 2026-05-14 rows link to

Trump, Donald J.-05.08.2026-278T.pdf and Trump, Donald J.-05.08.2026-278T(2).pdf.

Treat OGE as an official disclosure locator, not a clean holdings API. The PDFs are scanned/OCR'd forms; parse them carefully, preserve document dates and PDF URLs, and record whether each line is a purchase, sale, or bond/other instrument. Do not turn 278-T transaction ranges into exact position sizes or current holdings without a later confirming disclosure.

Scale check (correct the "$2B" myth)

• Q1 2026 disclosed trades: 3,600+ transactions, cumulative value

~$220M–$750M (range). He pivoted from mostly bonds (hundreds of millions, incl. ~$51M in March) toward hundreds of individual stocks.

• The "~$2 billion" figure people remember is **Trump Media's (DJT) ~$2B

Bitcoin / BTC-securities treasury*, not* his personal stock buying. Different thing.

• Net worth ~$6.3–6.6B (Forbes 2026), roughly doubled since 2024 — driven by

crypto + Trump Media paper gains, not his stock trades.

Disclosed 2026 stock buys (the watchlist)

Grouped by how "spent" the catalyst looks. Confirm price/news live before using.

Already had a Trump catalyst (catalyst largely spent — watch for round 2):

• DELL — Q1 buy + May-8 White House endorsement → popped to ATH, then $9.7B

Pentagon contract. The flagship example; further upside needs a new trigger.

Owns + plausible admin tailwind = top breakout candidates (1–2 month watch):

• PLTR (Palantir) — disclosed buy; admin steers government/defense contracts.

Catalyst to watch: a named federal contract or public praise.

• AXON (Axon) — disclosed buy; policing/border-security tailwind. Catalyst:

federal/border procurement, public mention.

• ORCL (Oracle) — disclosed buy; already a Stargate beneficiary, but more AI-

infra/federal-cloud announcements could re-trigger.

• NVDA, AVGO — disclosed buys; chips ride AI-infra announcements and any

export/tariff easing toward chips.

Owns, more macro/quiet (lower-conviction, theme-dependent):

• MSFT, META — mega-cap tech buys; move on broad AI/tariff-relief days.
• BAC, GS (financials) — move on deregulation / rate-cut pressure.
• Media basket: Paramount, WBD, NFLX, DIS, CMCSA — sensitive to his

praise/attacks and to FCC / merger policy he influences (M&A is the catalyst to watch). Note he has bought names he also publicly threatened.

• Bonds / munis — large but not catalyst-tradable here.

Personal / family holdings (mostly NOT things to chase)

• Trump Media & Technology Group (DJT) — his stake ~$1.3B (down from ~$4B

peak). Holds ~9,500+ BTC (~$0.6B) + Cronos (CRO). ⚠️ Pumps on his posts but is a long-term value destroyer: −~77% from its 2024 high; Q1 2026 net loss ~$406M on tiny revenue. Trade spikes, never hold as an investment.

• World Liberty Financial (WLFI) — family DeFi venture; ~22.5B WLFI tokens;

family cashed out ~$1.2B; Abu Dhabi-linked firm bought ~49% for ~$500M. Mostly private/token, not a clean equity trade.

• $TRUMP / $MELANIA memecoins — ⚠️ −97% / −99% from peak. Textbook

pump-and-dump; canonical "do not chase" cautionary tale.

• Personal ETH wallet — disclosed $1M–$5M.
• Blue chips also attributed: AAPL, MSFT, NVDA.

How to use this file

1. When a post names a company, check whether it's on this list → if yes, add the ownership amplifier (raises conviction; it's the Dell template). 2. For workflow (b) "build a watchlist", lead with the breakout-candidate group above and pair each with the specific catalyst that would trigger it. 3. When updating this file, start from the OGE search/API above, then use Reuters or issuer filings only as cross-checks or context. 4. Keep the structural caveat loud: ownership pumps are short-horizon; his own assets prove a Trump pump and a good investment are not the same thing.

Maintenance playbook (for the periodic timer)

This skill is meant to be run on a schedule (like the serenity skill) so the track-record.md ledger accumulates a real, measured hit-rate. Keep each run cheap and append-only; don't rewrite history.

Each scheduled run

1. Fetch the latest posts (Step 0):

   curl -sS -A "Mozilla/5.0" "https://trumpstruth.org/feed" -o /tmp/trump_feed.xml
   xreach tweets @realdonaldtrump -n 40 --json > /tmp/trump_x.json

Parse with the recipes in methodology.md. Truth Social is the main high- frequency text source; X is an official supplemental public channel and may be sparse, video-heavy, or repost-heavy. Process only new x:<tweet_id> candidates; if the latest X id has not changed, skip X classification quickly.

2. Dedupe by source id against the live ledger in track-record.md: use truth:<status_id> for Truth Social and x:<tweet_id> for X. Also cross- dedupe by normalized text or media title within a 24-hour window so a cross- posted message is not logged twice. Prefer the source with the clearest text; if both are equivalent, prefer Truth Social for continuity with the existing ledger and mention the X mirror in notes.

3. Classify each new post with the methodology.md checklist. skip anything not market-relevant (most of both feeds). Don't log skips except as a count. For X video/link-only posts, resolve the media/article/context before scoring; if context cannot be verified, store only a skip/count note in x-source-notes.md and do not append a prediction row.

4. Log a prediction for each market-relevant post: append a row to the live ledger (newest first, above the TIMER marker) with source id, date, source link, short text/context, tier + amplifiers, implicated ticker(s), and predicted direction / magnitude / durability. Leave T+1d / T+1w blank.

5. Score matured predictions. For rows whose post is now ≥1 trading day (and ≥1 week) old, fill in the actual move of the implicated ticker (use the project's market-data tools or a quick quote lookup) and mark hit/miss.

• "hit" = predicted direction realized over the stated horizon.
• Note separately whether a Tier-2 pop held a week or faded (durability is

the thing we most need to learn).

6. Recompute the rolling hit-rate table and update the calibration takeaways if a pattern's measured rate clearly diverges from the seed prior.

What earns a durable note vs. a one-off

Promote into the calibration takeaways only when it changes how we'd weight a pattern — e.g., "Tier 2C crypto posts hit direction 8/10 but held only 1/10", "named-company praise on a name he owns held 5/6". Don't editorialize single posts; let the counts talk.

X backfill queue

Use x-source-notes.md for historical X candidates. Do not bulk append old X items to the live ledger without scoring them. When backfilling:

1. Start with the three strongest candidates marked Backfill candidate. 2. Verify full text/media context. 3. Score T+1d and T+1w before inserting into track-record.md. 4. Preserve x:<tweet_id> as the source id and note any Truth Social mirror.

Hygiene

• Append-only ledger; never delete scored rows (the misses are the most valuable

calibration data).

• Keep provenance compact: source id + date + permalink, not full post text.
• If trumpstruth.org is down, continue with X if available and say the Truth

source failed. If X/xreach fails, continue with Truth Social and say X failed. If both fail, skip the run and retry next tick.

• Holdings change: refresh holdings-watchlist.md only when a new OGE

disclosure drops (roughly periodic), not every run. Use the official OGE Officials' Individual Disclosures Search Collection first: https://www.oge.gov/web/oge.nsf/Officials%20Individual%20Disclosures%20Search%20Collection?OpenForm. Its DataTables endpoint is https://extapps2.oge.gov/201/Presiden.nsf/API.xsp/v2/rest; filter the name column with Trump, sort docDate descending, then download direct PDF links from 278 Transaction rows for Trump, Donald J.

• OGE 278-T forms disclose transaction ranges, not exact current holdings. When

refreshing the watchlist, keep purchases, sales, bonds, ETFs, and family/private assets separate; never infer current position size from a transaction row.

Commit convention (if vendored into a repo like serenity)

• data: trump catalyst ledger update (+<n> posts) <UTC ISO> for ledger-only runs.
• skill: trump catalyst calibration update <UTC ISO> when takeaways/hit-rates

change.

• Don't create empty commits.

Methodology — what reliably pumps a stock, and how to read a post

The goal is to separate Trump posts that reliably push a stock up from the much larger pile of posts that don't move markets or that produce a spike which fades. Patterns are ranked by reliability of an immediate up-move, with an honest note on durability (does the pop hold, or is it a one-day trap).

All magnitudes are from documented cases (see track-record.md for sources). Treat them as rough bands, not guarantees — and remember a pop is a trade setup, not a fundamental endorsement.

---

Tier 1 — Names a specific stock (most reliable, immediate, single-name)

1A. Direct favorable mention of a specific public company. He names a company and praises it / tells people to buy / thanks the founders.

• Mechanism: retail + momentum front-run, and it often signals a coming policy

tailwind (contract, carve-out).

• Magnitude / durability: +5–15% intraday; can hold if a real catalyst

follows (a government contract, earnings), otherwise partial fade.

• Canonical case: Dell, 2026-05-08 — White House "go out and buy" → +~14.6%

intraday to an all-time high; weeks later a $9.7B Pentagon contract. He had also personally bought Dell in Q1. This is the template the whole skill watches for.

• Signal words: a company/ticker + great / tremendous / fantastic / the best

+ buy / support / go out and.

1B. Explicit all-caps "BUY" call (especially his own DJT or a clear ticker).

• Magnitude: +15–25%+ intraday; strongest when it's a stock he controls or

precedes a policy action he controls.

• Canonical case: "THIS IS A GREAT TIME TO BUY!!! DJT" → DJT +22.67% on the

day, and ~4 hours later he announced a 90-day tariff pause → S&P +9.5%.

• Caveat: draws explicit market-manipulation scrutiny. Durability is poor for

the named stock itself (DJT round-tripped); the broad-market leg held that day only because real policy followed.

Tier 1 is the "几乎确定会拉" bucket — but "pumps reliably" ≠ "good to hold."

---

Tier 2 — Policy / big-investment announcements (strong sector pop, mixed durability)

2A. White-House mega-investment announcement that names partner companies.

• Case: Stargate, 2025-01-21 — "largest AI infrastructure project in

history, $500B," on stage with Oracle/SoftBank/OpenAI → ORCL + AI datacenter / power names rallied; theme persisted for weeks.

• Signal words: $X billion, largest … in history, named CEOs/companies.
• Durability: medium — the theme outlives the day even when the headline name

cools.

2B. Tariff exemption / pause / relief for a named sector.

• Case: 2025-04-14 electronics exemption → AAPL +2.2%, DELL +4%; floated

auto-tariff pause → GM +3.5%, Ford +4.1%; broad relief rally.

• Durability: fragile — officials frequently walk it back ("temporary"),

so it can round-trip within days. Trade the reaction, distrust the follow-through.

• Signal words: exempt / exemption / pause / relief / carve-out + sector.

2C. Pro-crypto policy (strategic reserve, naming coins, anti-"Anti-Crypto Army").

• Case: 2025-03-02 strategic-reserve posts naming BTC/ETH/XRP/SOL/ADA → all

spiked Sunday, gave it back Monday. Drags COIN / MSTR / miners with it.

• Durability: lowest — typically a one-day spike. Good for a fast reaction

trade, dangerous to chase.

---

Tier 3 — Macro / thematic tailwinds (directional, not a single name)

These shift a whole basket rather than pumping one ticker; weaker and slower.

• "Drill baby drill" / energy dominance → oil & gas, nuclear/uranium, power.
• Pressuring the Fed to cut rates → small-caps, gold, long-duration growth.
• Defense / "peace through strength," or a conflict de-escalation → defense

primes, or broad risk-on.

• Reshoring / "Made in America" → domestic manufacturers, US steel, semis.

See theme-ticker-map.md for the specific names per theme.

---

Amplifier features — the more that stack, the higher the hit-rate

Score each market-relevant post on these. Several together ≈ the "几乎确定" zone: 1. Names a specific public company (not a vague sector) → directly tradable. 2. Superlative + call to action (great + go buy). 3. Precedes a policy action he controls (tariff move, contract, EO) → can self-fulfill. 4. He / his family personally owns it (cross-check holdings-watchlist.md) → the Dell template; ownership + an admin tailwind is the strongest combo. 5. Posted intraday or pre-open → immediate spike (after-hours leaks to next day). 6. Repeated / multi-post on the same name or theme → conviction.

---

Anti-patterns — do NOT read these as reliable up-signals

• Pure geopolitics rants with no named beneficiary (e.g., the steady stream of

Iran/foreign-policy posts) → diffuse, unpredictable market impact.

• Crypto-reserve / coin posts → usually one-day spikes; don't chase late.
• Tariff "relief" that gets walked back the same week → false breakout.
• Attacks on a company/industry → these dump, not pump (opposite direction);

treat as a short-side or avoid signal, not a buy.

• His own promotional posts about DJT / memecoins → may spike, but these are

structurally value-destroying for holders (DJT −77% from 2024 high; $TRUMP −97%; $MELANIA −99%). Flag the trap.

---

Per-post classification checklist

For each fresh post, answer in order — stop early if it's not market-relevant:

1. Market-relevant? Mentions a company/ticker, sector, tariff/trade, crypto, energy, the Fed/rates, defense, AI/infra, or a $-investment? If no → skip. 2. Tier? 1 (named stock) / 2 (policy-sector) / 3 (macro theme). 3. Direction? Favorable (pump) vs. attack (dump). 4. Amplifiers present? Count them (list above). 5. Ownership overlap? Cross-check holdings-watchlist.md. 6. Implicated tickers via theme-ticker-map.md. 7. Reliability + durability call: combine Tier + amplifiers + the historical base-rate in track-record.md → state likely magnitude and whether it can hold or is a fade. 8. Log it for the timer (see maintenance.md): source id, date, source, classification, predicted direction/magnitude — so the outcome can be scored later.

---

Fetch + parse recipe (Step 0)

Truth Social via trumpstruth.org RSS

curl -sS -A "Mozilla/5.0" "https://trumpstruth.org/feed" -o /tmp/trump_feed.xml

# Newest-first list of recent posts: (utc, status_id, permalink, text)
import xml.etree.ElementTree as ET, html, re
items = ET.parse('/tmp/trump_feed.xml').getroot().find('channel').findall('item')
def clean(s): return re.sub(r'\s+', ' ', html.unescape(re.sub('<[^>]+>', ' ', s or ''))).strip()
posts = []
for it in items:
    link = it.findtext('link') or ''
    sid  = (re.search(r'/statuses/(\d+)', link) or [None, None])[1]
    posts.append({
        'source_id': f'truth:{sid}',
        'source': 'Truth Social',
        'date': it.findtext('pubDate'),
        'link': link,
        'text': clean(it.findtext('description')),
    })
# posts[0] is the most recent. Filter to market-relevant via the checklist above.

Notes:

• The feed is a rolling window (~100 posts / ~5 days), newest-first. For the timer,

dedupe by `truth:<status_id>` so each post is scored once.

• RT: items are reposts; resolve the linked status if it matters.
• The official truthsocial.com API is Cloudflare-blocked from servers. If you

ever need full fields / deeper history, truthbrush (needs a Truth Social login) or a real browser (agent-browser) are the fallbacks — but for catalyst spotting this RSS is enough.

X via @realDonaldTrump

xreach tweets @realdonaldtrump -n 40 --json > /tmp/trump_x.json

import json
data = json.load(open('/tmp/trump_x.json'))
items = data.get('items') if isinstance(data, dict) else data
posts = []
for t in items or []:
    tid = str(t.get('id') or '')
    if not tid:
        continue
    posts.append({
        'source_id': f'x:{tid}',
        'source': 'X',
        'date': t.get('createdAt'),
        'link': f"https://x.com/realdonaldtrump/status/{tid}",
        'text': t.get('text') or '',
        'is_retweet': bool(t.get('isRetweet')),
        'is_quote': bool(t.get('isQuote')),
        'media': t.get('media') or [],
    })

X-specific caveats:

• Treat X as an official supplemental channel, not a replacement for Truth

Social. It can be sparse and may contain video-only links or retweets.

• For video/link-only posts, resolve the media, linked article, or visible

caption before scoring. If the market-relevant claim cannot be verified, skip rather than logging a prediction.

• Retweets without Trump commentary generally count as lower-conviction leads.
• Cross-posted Truth Social/X messages should be logged once, using the clearest

source and noting the mirror source.

Theme → ticker map

When a post is Tier 2/3 (a policy or theme rather than a single named company), use this to translate the topic into the names that typically react. These are candidate reaction baskets, not recommendations — always confirm the current setup and check track-record.md for how durable that theme's reactions are.

Direction assumes a favorable Trump post. If he attacks the theme/company

instead, flip the sign — and remember attacks tend to dump harder than praise

pumps.

Crypto / digital assets

• Trigger words: bitcoin, crypto, strategic reserve, "Anti-Crypto Army",

stablecoin, CFTC, prediction markets, World Liberty, $TRUMP.

• Reacts up: COIN, MSTR, HOOD; miners MARA/RIOT/CIFR/CLSK; his own DJT

(holds BTC). Coins: BTC, ETH, SOL, XRP.

• Durability: ⚠️ low — usually a one-day spike that fades. Fast reaction trade.

AI infrastructure / data centers

• Trigger words: Stargate, AI infrastructure, "largest in history", data

centers, chips, compute, electricity for AI.

• Reacts up: ORCL, NVDA, AVGO, PLTR, DELL, SMCI; power/grid CEG, VST, GEV,

NRG; he disclosed owning ORCL, NVDA, AVGO, PLTR.

• Durability: medium — theme persists past the headline.

Tariffs / trade

• Exemption / pause / relief (UP): the relieved sector rallies. Electronics →

AAPL, DELL, HPQ; autos → GM, F, TSLA; semis → NVDA, AVGO, broad chips.

• New / higher tariffs (DOWN): importers, retailers, autos, China-exposed

names fall; domestic substitutes may rise.

• Durability: ⚠️ fragile — frequently walked back within days.

Energy / power

• Trigger words: "drill baby drill", energy dominance, LNG, nuclear, grid.
• Reacts up: XOM, CVX, OXY, COP; LNG, ET, KMI; nuclear/uranium CEG, VST,

CCJ, OKLO, SMR; coal occasionally.

Defense / space / security

• Trigger words: peace through strength, Golden Dome, military buildup,

drones, border.

• Reacts up: LMT, RTX, NOC, GD, LHX; PLTR, AXON; space RKLB, LUNR.
• De-escalation / "deal done": broad risk-on; oil can fall (supply fear off).

Reshoring / "Made in America" / manufacturing

• Reacts up: US steel (NUE, STLD, CLF, X), domestic semis (INTC, GFS, TXN),

caterpillar-type industrials (CAT, DE), shipbuilders.

Pharma / healthcare

• Drug-price controls / "most favored nation" (DOWN): large-cap pharma

(PFE, MRK, LLY, JNJ) tends to fall.

• Deregulation / FDA-friendly (UP): biotech/XBI risk-on.

Media / "his" companies & conflict-of-interest overlaps

• DJT (Trump Media): spikes on any DJT-favorable post; ⚠️ structurally weak —

trade the spike, do not hold.

• Disclosed media buys (Paramount, WBD, NFLX, DIS, CMCSA): sensitive to

Trump praise/attacks and to FCC/merger policy he influences.

Financials

• Deregulation / rate-cut pressure (UP): banks BAC, GS, JPM (he disclosed

BAC, GS); small-caps IWM; rate-cut talk → gold (GLD, GDX), long-duration growth.

---

Names in bold ownership overlaps (ORCL, NVDA, AVGO, PLTR, DELL, BAC, GS, media basket) carry the extra "he owns it" amplifier — see holdings-watchlist.md. Ownership + a controllable admin tailwind is the highest-conviction setup in this skill (the Dell template).

Trump Catalyst Track Record

Live ledger of market-relevant posts from @realDonaldTrump (Truth Social primary, X secondary). Newest entries first. Leave T+1d / T+1w blank when posting; fill in after the relevant trading day(s).

Tier guide:

- Tier 1 (High) — explicitly names a company he owns + praises it / announces a deal. Dell template.

- Tier 2 (Medium) — policy tailwind for a sector; or named but not clearly owned; or owned but indirectly implied.

- Tier 3 (Low) — macro/sentiment/political; no named company or direct sector call.

<!-- TIMER -->

@realDonaldTrump X source notes

Purpose

Use this file to preserve source-specific observations for Donald Trump's X account, https://x.com/realdonaldtrump, without overloading the main prediction ledger.

X is an official public Trump channel, but it behaves differently from Truth Social:

• lower-frequency than Truth Social in the current setup;
• often video-only, short-link, or repost-heavy;
• high engagement can amplify a message, but engagement alone is not a market

signal;

• context must be resolved before a video/link-only post is scored;
• cross-posted Truth Social/X messages should be logged once in

track-record.md, with the mirror source noted.

Initial maintenance rule

For every scheduled run:

1. Fetch latest X posts with xreach tweets @realdonaldtrump -n 40 --json. 2. Normalize each candidate as x:<tweet_id>. 3. Cross-dedupe against Truth Social ledger rows using source id, normalized text, media/link context, and a 24-hour window. 4. Append to the market ledger only when the X item has verified context and a tradable market read. 5. Keep video/link-only or retweet items here as source notes unless the market context is independently verified.

Historical X seed candidates

The historical seed table is intentionally compact. It is for candidates that may be worth adding to track-record.md or citing as source-behavior examples.

Initial partial backfill, captured with xreach tweets @realdonaldtrump --json --count 100 --all --max-pages 12 --delay 1200:

• Captured 240 X timeline items.
• Coverage was incomplete (complete: false).
• Range: 2024-11-06 00:08:02 UTC through 2026-05-22 23:34:07 UTC.
• Shape: 22 retweets, 9 quote posts, 0 replies, 138 media posts, 68 videos,

72 images, 174 posts containing t.co, and 114 posts whose visible text was only a short link. Many video/link-only posts are context-unverified.

Context-unverified / skip examples

Maintenance recommendation from partial backfill

X is useful as a supplemental official channel and historical backfill source, but it should not displace Truth Social as the main feed. The timeline capture was incomplete and many posts were video/link-only.

• Fetch X on scheduled runs, but process only new x:<tweet_id> candidates and

skip quickly when the latest id has not changed.

• Candidate keywords: named company, explicit buy/praise, tariff, crypto,

energy, healthcare, Fed/rates, shipping/oil, geopolitical closure/blockade, and large fiscal/tax bills.

• Treat retweets and quote posts as lower-confidence unless Trump adds original

market-relevant commentary.

• Use X backfill mainly to seed 2024-2025 patterns, cross-check Truth Social,

and capture rare X-only named-company posts.