Built AISH to make terminal output more useful without changing how commands run. Your agent context will love it.

What it does:

  - Runs commands in a real PTY (colors/progress still work)
  - Logs full output for every run
  - Shows short digest on success
  - Shows relevant excerpt on failure (tool-aware detectors + generic fallback)
  - Supports bash/zsh via ai wrapper and optional shims

Example:

  - aish-run -- cargo test
  - aish-run --last
  - aish-run --open

Current detectors include pytest/jest/vitest/cargo/go/tsc/eslint/ruff/mypy/maven/gradle/dotnet/cmake/terraform/docker/kubectl.

Would love feedback on:

  - detector quality / false positives
  - shell UX defaults
  - packaging/release experience