Back to skills
Skill

gad:eval-bootstrap

gad-eval-bootstrap

Bootstrap an eval agent with full GAD context injected into its prompt

Generate a bootstrap prompt for a single eval project and launch an agent in an isolated worktree. The agent starts with all GAD context inline — it does not need to discover the loop.

  1. Generate the bootstrap prompt:

    gad eval run --project $PROJECT --prompt-only

  2. Read the generated PROMPT.md from the eval run directory.

  3. Launch an agent with isolation: "worktree" using the prompt content.

  4. After completion, reconstruct the trace:

    gad eval trace reconstruct --project $PROJECT
Read and follow the companion `eval-bootstrap` skill if it is installed.
Source on GitHub

vendor/get-anything-done/sdk/skills/gad-eval-bootstrap/SKILL.md

Client debug · NEXT_PUBLIC_CLIENT_DEBUG=1
0 lines

No events yet. Window errors, unhandled rejections, and React render errors appear here. Set NEXT_PUBLIC_CLIENT_DEBUG_CONSOLE=1 to mirror console.error / console.warn.