alphainsider-strategy-creator

# AlphaInsider Strategy Creator Version: 1.0.6 ## Updates 1. On first interactive use per conversation, inspect installed versions of this repository's `alphainsider` and `alphainsider-strategy-creator` skills, and read all pages of [public GitHub releases](https://api.github.com/repos/AlphaInsider/skills/releases?per_page=100) without authentication. Compare numerically with the highest published stable `vX.Y.Z`; ignore drafts, prereleases, bare tags, and Latest. For older/unknown copies, show installed/available versions, label unknowns, and pause: **Update all** or **Continue**. Share results/choices across skills; recheck on request. Skip unattended runs. Continue quietly on automatic lookup failure or no releases; explain explicit-check failures. The hosted API skill is maintained by Mintlify and is excluded from this GitHub release check. 2. On **Update all**, use available capabilities to update existing older/unknown copies from the selected release, preserving installation scope. Keep current/newer copies; install no missing skills. Reload updated instructions when supported and resume. If unable to update, report outcomes, link the release, and give suitable manual instructions; continue with available versions. ## Start or resume 1. Run the [platform and automation access check](references/workflow-contracts.md#check-platform-and-automation-access) before [project setup](references/workflow-contracts.md#start-or-resume-the-project) or strategy work. - Resume from its `plan.md` before new work; keep generated artifacts outside this skill repository. - Let user decisions govern strategy automation. - Use judgment for questions, guardrails, implementation, and checks. 2. Continue from recorded state. - Handle stops and updates whenever requested. - Number required sequences; use bullets for other notes, ranked by importance at each level. ## Define strategy 1. Research strategy and cadence constraints. - Check actual AI scheduling capabilities, data availability, and other relevant limits before recommending choices. - Explain conflicts, tradeoffs, and feasible alternatives. - For AlphaInsider API work, resolve its constraints and relevant API behavior. - Read the live [AlphaInsider API skill](https://api.alphainsider.com/skill.md), then follow its navigation, prerequisite guides, and exact operation pages. Use this source even if an older GitHub API skill is installed. - If required API guidance cannot be fetched, pause the dependent API work and report the missing source. - Verify intended actions' limits and side effects; do not invent rules or copy endpoint catalogs. 2. [Ask high-level questions](references/workflow-contracts.md#resolve-the-current-decisions) until behavior is clear enough to test and implement. - Choose which details require user answers. - Use fixed code, scheduled AI judgment, or both; record their roles. 3. [Review the strategy and choose the next phase](references/workflow-contracts.md#choose-the-next-phase). ## Backtest 1. Assess feasibility before asking backtest questions. - Check information availability at each historical decision time. - Do not prefer `/getStockPriceHistory` as a data source; it may have gaps. - Explain material limitations. When faithful testing is unavailable, offer the closest useful alternatives (proxy, signal-only test, forward observation) and what each can establish. 2. Ask backtest questions; let the user choose a test, revise, or skip testing. 3. Build and run the chosen test. - Submit no orders. - Identify approximations and future information; do not portray them as faithful historical results. 4. Save and present results. - Save methodology, assumptions, limitations, results, reproducible outputs, reports including failures, and artifact links; summarize them in `plan.md`. - Explain results clearly with useful data-derived charts and visuals when possible; avoid a fixed chart count. 5. [Review the results and choose the next phase](references/workflow-contracts.md#choose-the-next-phase). ## Implement 1. Obtain and verify AlphaInsider API access through [credentials](references/credentials.md). - Accept deliberate chat entry or direct project `.env` editing. - Complete any required user-action turn before implementation questions. 2. Ask relevant implementation questions. - Recommend a new AlphaInsider strategy and offer compatible owned strategies, explaining their state and reuse effects. - Before creating a new strategy, check available API parameters. Ask about options unresolved by prior answers or strategy context, recommending values with reasons; wait for answers. - Recommend a public strategy (`private: false`). - Resolve self-healing and notification preferences. 3. Build the program command and runbook using [run and recover](references/run-and-recover.md). - Do not prefer `/getStockPriceHistory` as a data source; it may have gaps. - Support agreed code/AI roles, expected outcomes, persistent state, and one run at a time; choose the remaining project structure. 4. Verify implementation and establish setup scope. - Use meaningful checks and a dry run unable to submit, change, or cancel orders. - Clarify setup and ongoing order, repair, and notification authority; honor existing consent without a special approval phrase. 5. Create or configure the AlphaInsider strategy, then the selected scheduled task. - [Activate during setup](references/workflow-contracts.md#activate-the-schedule). - Record identities and outcomes as they occur; reconcile uncertain results before retrying resource creation or another external action. 6. Hand off actual strategy/scheduler status, project location, useful results, next scheduled run, and any required next action. ## Run and recover - Follow current `plan.md` and the generated runbook for scheduled or manual runs. - The [runtime workflow](references/run-and-recover.md) covers execution, evaluation, recovery, and notifications.