haindy

Use when setting up a fresh HAINDY installation, configuring API credentials for the first time, migrating from a legacy .env file, or verifying the effective configuration.

Use before release-facing, provider, runtime, or surface changes to manually regression-test HAINDY through its own public tool-call commands across configured providers and available surfaces.

Investigate failed HAINDY batch or tool-call executions by reconstructing events from the effective data root, model logs, reports, and tool-call session artifacts under `~/.haindy/sessions/<session-id>/`. Use when a user asks for forensics on the last or a specific HAINDY run/session, wants a high-level failure timeline, issue analysis, or candidate fixes from local HAINDY artifacts. Do not use this skill for ordinary `pytest` stack-trace debugging.

Use when setting up a fresh HAINDY installation, configuring API credentials for the first time, migrating from a legacy .env file, or verifying the effective configuration.

Use when preparing, validating, or publishing a HAINDY release from this repository, bumping package versions, updating the changelog or release-facing docs, or verifying the tag-driven PyPI and GitHub release flow. Covers required local validation, release commit and tag steps, and HAINDY-specific version guardrails such as keeping `haindy version` wired to installed package metadata instead of a hardcoded string.

Use when you need to drive HAINDY's tool-call mode to inspect a live desktop, Android, or iOS session, execute direct UI actions, run scenario-based tests, manage session variables, or interpret HAINDY's JSON command results.