agentleFS
Sign inStart free
YichengYang-EthanYichengYang-Ethan/oracle3/skills/pm-agent-strategy-authoring/SKILL.md

pm-agent-strategy-authoring

Skill466 starsChanged 34 days agoApache-2.0View on GitHub

Turn an LLM- or tool-driven idea into AgentStrategy code and evaluate it with paper trading.

We split this skill into the rules it gives an agent

8 rules. About 0 are common, 4 are distinctive, and 4 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈4

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 4

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Use this skill when the user asks for a strategy that makes decisions with an LLM, MCP tools or external APIs.

    line 8
  2. Distinctive:

    No parameter grid search: agent strategies cannot be auto-tuned.

    line 14
  3. Very specific:

    Call the LLM, MCP tools or external APIs inside process_event.

    line 30
  4. Very specific:

    Write the LLM output into the reasoning field with self.record_decision(reasoning=...) so it can be monitored.

    line 32
  5. Very specific:

    Call trader.place_order(...) to act.

    line 33
  6. Very specific:

    Check self.is_paused() before every decision.

    line 55
  7. Distinctive:

    Never use future information (no look-ahead).

    line 56
  8. Distinctive:

    Keep runs reproducible: record the strategy file path, class name and command.

    line 57

≈ “Common” is an estimate: at least 1 in 5 of the 308,232 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

PM Agent Strategy Authoring

Use this skill when the user asks for a strategy that makes decisions with an LLM, MCP tools or external APIs.

Goals

  • A runnable strategy class (an AgentStrategy subclass).
  • strategy validate and paper run both succeed.
  • No parameter grid search: agent strategies cannot be auto-tuned.

Code entry points

  • Base class contract: oracle3/strategy/agent_strategy.py (AgentStrategy).
  • LLM strategy reference: oracle3/strategy/simple_strategy.py.
  • New strategies go in strategies/.

Steps

  1. Create the skeleton
  • oracle3 strategy create --output strategies/<name>.py --class-name <ClassName> --type agent
  1. Implement the strategy
  • Call the LLM, MCP tools or external APIs inside process_event.
  • Non-determinism is fine: results may differ between calls.
  • Write the LLM output into the reasoning field with self.record_decision(reasoning=...) so it can be monitored.
  • Call trader.place_order(...) to act.
  • Guard every decision with self.is_paused() to respect the control plane.
  1. Quick validation
  • oracle3 strategy validate --strategy-ref strategies/<name>.py:<ClassName> --dry-run --events 10 --json
  1. Evaluate with paper trading (the right way to evaluate agent strategies)
  • oracle3 paper run --strategy-ref strategies/<name>.py:<ClassName> --monitor
  • Watch decisions and positions with oracle3 trade status --json.
  • Take a full snapshot with oracle3 trade state --json.
  1. Intervene and adjust
  • oracle3 trade pause / oracle3 trade resume
  • Hot-swap the strategy with oracle3 trade swap --strategy-ref strategies/<new>.py:<NewClass>.

Hard rules

  • AgentStrategy subclasses cannot be used with research auto-tune --param-grid-json; it raises an error.
  • External API calls need timeouts and error handling so they never block the event loop.
  • Check self.is_paused() before every decision.
  • Never use future information (no look-ahead).
  • Keep runs reproducible: record the strategy file path, class name and command.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

Check your setup, free