Stock Direction Signals

stock-direction-signals

Directional research on one US equity ticker or index proxy. The skill gathers paid signal from a fixed provider filter — Alpha Vantage MPP (quote, chart, RSI, MACD, news sentiment, earnings), SELAT-native Twitter (raw chatter), Circle StableEnrich (Reddit threads), and Circle Otto (TradFi macro regime) — and the agent fuses it into one brief: verdict (bullish / bearish / mixed / insufficient data), confidence, the technical setup, catalysts, social sentiment, macro backdrop, and the contrarian risks that would invalidate the read. Research only — never trade execution or financial advice.

When To Use

Use when a user asks whether a ticker looks bullish, bearish, or mixed over a short or medium horizon and the answer needs live market context plus social or news sentiment — MAG7 names, semiconductor and AI-infrastructure stocks, tokenized-stock watchlists, or index proxies like SPY and QQQ.

Do not use for trade execution, portfolio advice, or any request that should place an order. This skill only gathers paid data and synthesizes a research brief; if the user asks to trade on the result, decline that part and say so.

Rails

This skill spans three settlement modes, so the skill's rail is mixed:

• x402 via Circle Gateway — SELAT-native Twitter (catalog.selat.ai) settles x402 via Circle Gateway through the SELAT Router.
• x402 on Base — Circle Otto (x402.ottoai.services) settles x402 on Base through the SELAT Router.
• MPP on Tempo — the six Alpha Vantage steps (alphavantage.mpp.paywithlocus.com) and Circle StableEnrich (stableenrich.dev) settle MPP on Tempo through the SELAT Router.

The selat CLI auto-detects each step's protocol and settlement mode at call time. The runner pays on whichever chain holds Gateway USDC — rail names are not a pay-chain claim. A reachable SELAT Router is required.

Workflow

1. Install: selat skill install stock-direction-signals
2. Run end-to-end: selat skill run stock-direction-signals --ticker <TICKER> --twitter_query "<query>" --reddit_query "<query>"
3. The CLI compiles each step into a selat-pay call and prints each result.

Before running, tell the user what it costs: a full run quotes at roughly $0.08 live (nine paid calls, capped at $0.25 total) — say "this pulls nine paid data feeds for about eight cents — proceed?" and wait for a yes before spending. Afterwards, report what was actually spent.

Recommended agent procedure (manifest steps are ordered cheapest-first; retarget all three params to the user's ticker — never let the NVDA defaults run for a different company):

1. Normalize the request into ticker, twitter_query, reddit_query, and a horizon (intraday / week / month / quarter). Preserve the user's ticker exactly; expand obvious company names into their tickers. Fold company and product names into the two query params (e.g. $NVDA OR NVIDIA OR Blackwell).
2. Social pulse — SELAT-native Twitter advanced search (~$0.001). Raw evidence only: the agent scores tone, intensity, recency, and engagement quality itself.
3. Macro regime — Circle Otto tradfi-data (~$0.001–0.005): indices, VIX, DXY, yields, commodities, and quote context for risk-on/risk-off framing.
4. Market core — Alpha Vantage global-quote, time-series-daily, RSI, MACD (~$0.008 each): last price, chart trend, momentum, and trend acceleration.
5. Catalysts — Alpha Vantage news-sentiment and earnings (~$0.008 each): news tone, catalyst clustering, earnings surprise history.
6. Retail texture — Circle StableEnrich Reddit search (~$0.02): community debate and long-form reactions; weight recency, comment quality, and claims repeated across communities.
7. Synthesize a compact brief: verdict, confidence, price and technical setup, catalysts and fundamentals, social sentiment, macro/regime, contrarian risks, and source notes. Label it bullish, bearish, mixed, or insufficient data — and say what would invalidate it.

Relay the brief in plain language with the dollar cost of the run; keep endpoint URLs and raw response JSON out of what the user sees. When the user only wants a subset (e.g. no social), run the individual steps with selat-pay instead of the full manifest. For multiple tickers, repeat the run per ticker.

For deeper dives, the provider filter also offers optional endpoints not in the manifest (intraday/weekly/monthly series, SMA/EMA/BBANDS, earnings-call transcripts, financial statements, tweet replies, Reddit post comments, Otto derivatives context) — see references/endpoints.md; call them individually via selat-pay within the same provider filter.

Inputs And Outputs

Output: per-step JSON (tweets, macro data, quote, OHLCV series, RSI, MACD, news sentiment, earnings history, Reddit posts) that the agent fuses into a non-advisory directional brief — what moved the signal, what would invalidate it, and which paid endpoints were used.

Gotchas

• The provider filter is the skill. Use only Alpha Vantage MPP, SELAT-native Twitter, Circle StableEnrich, and Circle Otto. Do not substitute AIsa, BlockRun, Serper, Apollo, Exa, CoinGecko, Nansen, or other catalogue matches even if they rank higher.
• Retarget the defaults. All three params default to NVDA-flavored values so verify can exercise every step; a run for any other ticker must override all three or the social steps research the wrong company.
• Alpha Vantage steps are POST — params go in body, keyed by symbol (indicators also need interval/series_type, RSI adds time_period). news-sentiment keys on tickers (plural) — and its topics field takes a fixed topic enum, not a company name, so this manifest omits it.
• SELAT-native Twitter returns raw social objects. Never claim a finished sentiment score unless the response explicitly includes one; the scoring is the agent's job.
• Reddit is noisy and momentum-skewed. Use it for thesis texture; weight recency, comment quality, and repetition across communities.
• Otto is macro-first here. tradfi-data is the default; use Otto's crypto/perp endpoints only when the user explicitly connects the stock to tokenized equity, crypto beta, or derivatives spillover.
• maxAmount is a guardrail, not the price. Per-step caps ($0.01–$0.04, $0.25 full-run) sit well above live quotes (measured full run ≈ $0.0756); gateway prices can run above the catalogue listing.
• The live 402 is the source of truth. If a step stops serving a challenge or the live price drifts, selat skill verify flags it — report the live response and adapt within the provider filter only.
• Non-advisory, always. The brief can support a bullish or bearish thesis but must never be presented as financial advice or a guaranteed prediction.

Validation

--chain base in the probe command below is only the flag selat-pay requires today — a probe reads a free, chain-independent quote and never settles. A real paid run resolves the settlement chain from your funded Circle Gateway balance, not the manifest.

• Static: selat skill validate ./skills/stock-direction-signals
• Live gate (free): selat skill verify ./skills/stock-direction-signals --ticker NVDA --twitter_query "\$NVDA OR NVIDIA" --reddit_query "NVDA stock"
• Paid confirm (settles real 200s): add --pay to the verify command.
• Single-step probe (no pay): selat-pay POST "https://alphavantage.mpp.paywithlocus.com/alphavantage/global-quote" --body '{"symbol":"NVDA"}' --chain base --probe-only

References

• manifest.json — the machine-readable payment recipe this skill runs.
• references/endpoints.md — the provider-filtered catalogue endpoints, request shapes, and prices.
• references/agent-skill-authoring-sop.md — authoring standard.
• selat-pay — https://github.com/SELAT-AI/selat-pay