Trading Brain
WarnAudited by ClawScan on May 10, 2026.
Overview
This instruction-only skill is not clearly malicious, but it should be reviewed because it can steer brokerage-related trading decisions, use local account configuration, and persist trade notes without clear approval boundaries.
Before installing, decide whether you want an agent to influence real trading decisions. Keep this skill analysis-only unless you explicitly approve each trade, verify the local Alpaca/trade-bot configuration is read-only where possible, review the referenced Serper helper script, and monitor the memory files it writes.
Findings (4)
Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.
If connected to trading tools, the agent could make or strongly steer financial decisions that may cause losses without the user explicitly approving each action.
The skill is meant to guide concrete trading decisions, including a recurring cron context and dollar-sized trades, but it does not define a required user approval step before high-impact financial actions.
Starting a trading cron ... Making a buy/sell/hold decision ... TQQQ/SQQQ Daily Direction - $5K per trade on Account 2 - Only trade if conviction >55%
Require explicit user confirmation before any order placement, account change, or trade execution, and keep automated cron use limited to analysis/reporting unless separately authorized.
The agent may read brokerage account state or use configuration that could contain sensitive account settings or credentials.
The skill directs the agent to use a local brokerage/trade-bot configuration to update account positions, introducing account-level access that is not clearly bounded in the metadata.
Current Positions (update from Alpaca before deciding) Config: `~/Documents/GitHub/trade-bot/config.yaml` Account ID: Account 1 (manual trading)
Confirm exactly what the Alpaca config contains, restrict it to read-only position access where possible, and avoid giving the agent trading credentials unless explicitly intended.
The actual behavior of the referenced local search script depends on code outside this skill.
The skill depends on a local shell helper that is not included in the one-file artifact set. This appears purpose-aligned for news search, but the helper itself was not reviewed here.
Serper (Google News) — `~/clawd/integrations/serper/search.sh "query" 5 news` — USE THIS FIRST
Review the local `serper/search.sh` script and ensure it only sends intended search queries to the expected provider.
Incorrect, outdated, or manipulated memory entries could influence later investment analysis or recommendations.
The skill intentionally persists trade history and strategy updates in local memory files, which is aligned with its purpose but can affect future trading analysis if stale or poisoned.
Document every trade: thesis, outcome, lesson learned → `~/clawd/memory/trades.md` ... Update candidates in `~/clawd/memory/trading/TRADING-BRAIN.md`
Periodically review these memory files, keep secrets out of them, and require user approval before using persisted notes as a basis for trades.