AI CLI Bootstrap Skill

AI CLI Bootstrap Skill

Overview

Use this skill when direct access is the only blocker. Reuse an existing local tool first, otherwise generate the smallest task-scoped CLI needed to fetch or transform the missing artifact, run it immediately, and answer from the result.

Workflow

1. Identify the missing artifact and the final answer format.
2. Reuse an existing local script, binary, or integration if one already solves the gap.
3. If nothing suitable exists, scaffold a helper in .agent-tools/<slug>/cli.py.
4. Keep the helper agent-facing:
   • minimal arguments
   • plain text or JSON output
   • no interactive prompts
   • environment variables for secrets
5. Run the helper on a real or representative sample before trusting it.
6. Use the helper output to finish the original task.

Scaffold

Generate a starter helper with:

python scripts/bootstrap_ai_cli.py youtube-script \
  --purpose "Fetch transcript text and metadata for a YouTube video" \
  --template transcript_stub \
  --workspace "$PWD"

If the host repository already uses another scratch-tool convention, override it with --tools-dir.

Rules

• Prefer a single Python file.
• Prefer the standard library first.
• Implement only the missing edge.
• Stop calling it a tiny CLI if it starts needing config, subcommands, or framework structure.

References

• references/ai-only-cli-rules.md: detailed operating rules
• references/decision-guide.md: build vs reuse vs stop
• references/anti-patterns.md: what not to build
• templates/: starter scaffolds by task type