Current Equity Research Pipeline

Current scheduled job

Page generated: 2026-06-02T08:36:41.467424+00:00 UTC. Research support only; not financial advice.

name

Equity X signal + agentic tool brief (8h)

id

a1c48b2edb9d

schedule

every 480m

enabled

True

state

scheduled

last_run_at

2026-06-02T08:11:48.138697+00:00

last_status

ok

next_run_at

2026-06-02T16:11:48.138697+00:00

script

equity_signal_cron_runner.py

deliver

origin

enabled_toolsets

[
  "terminal",
  "file",
  "web",
  "skills"
]

skills

[
  "equity-signal-research",
  "hermes-cron-signal-briefs"
]

Pipeline stages

Skills and agents used

Loaded / referenced skills

equity-signal-researchhermes-cron-signal-briefssector-overviewcompetitive-analysiscomps-analysisearnings-analysismodel-updatedcf-model

Agent layers

hermes_cron_formatter

LLM pass that converts deterministic script JSON into a concise Telegram brief.

ai_hedge_fund

[
  "fundamentals_analyst",
  "valuation_analyst",
  "technical_analyst",
  "sentiment_analyst"
]

tradingagents

[
  "StockTwits social lens",
  "Reddit/RSS social-news lens"
]

citation_verifier_team

[
  "sec-filing-and-facts",
  "market-and-social-sources",
  "citation-coverage-and-format"
]

Prompts used by the running cron

Hermes Telegram brief prompt

You are running the equityanalysis recurring signal brief. A deterministic pre-run script has already collected X/Twitter/RSS ticker signals, run bounded third-party agentic tool gates, and emitted JSON into the prompt context.

Format a concise Telegram brief for Jeremy. Rules:
- Treat X/Twitter/RSS/social posts as discovery only, not investment merit.
- Do not give financial advice, price targets, buy/sell/hold instructions, or trade recommendations.
- If collector_status or return_code indicates failure/no data, report the failure succinctly and do not invent signals.
- Use only tickers and evidence present in the script JSON.
- Promote only tickers with cross-account evidence or explicit script-selected status.
- Summarize ai-hedge-fund as a readiness/planned rubric pre-screen unless the script reports real executed results; do not imply AIHF ran if it was skipped.
- Summarize TradingAgents only as a selective StockTwits/Reddit social/news lens; mention unavailable/degraded sources when reported.
- If auto_report_result contains published/verified report URLs, include them. If no reports were generated, say no new full report was generated.
- Be sparse: if there are no strong signals, say so in 2-4 bullets.

Output format:
## Equity signal brief
- Generated: <timestamp>
- Collection: <runtime mode/status>

## Top signals
For up to 5 selected tickers:
- **TICKER**: score/mentions/accounts; one-line why surfaced; include 1 source URL if present.

## Agentic tool layer
- AIHF: executed/ready/skipped + reason; planned tickers if skipped.
- TradingAgents social lens: tickers run + one useful StockTwits/Reddit observation if grounded in the JSON.

## Reports
- New full reports/deploy status if present; otherwise no new full report.

End with: Research support only; not financial advice.

Citation verifier prompt roles

{
  "common": "Fresh context citation verifier subagent. Inspect the report and structured generation data. Fetch or inspect cited URLs where possible. Pass only if citations support claims, labels/dates are honest, and calculations tie to cited public data.",
  "roles": [
    "sec-filing-and-facts",
    "market-and-social-sources",
    "citation-coverage-and-format"
  ]
}

Machine-readable snapshot

The same method state is published as JSON for quick inspection: data/pipeline/current.json.