Puter Deployer

Use a strict flow: preflight → build/stage → deploy path selection → verify → rollback note.

1) Preflight (always)

Run:

bash skills/puter-deployer/scripts/preflight.sh <project_dir> <build_dir>

This checks:

• puter CLI exists
• puter whoami works
• build directory exists and is non-empty
• index.html exists

If any check fails, stop and report exact fix commands.

2) Deployment path selection

Choose one path explicitly:

1. Create new Puter app entry (CLI-supported):
   • puter app:create <name> <remoteDir> --description "..." --url "..."
2. Update/publish build content:
   • Use Puter API fallback guidance in references/api-fallback.md.
   • If API details are unclear for this host/version, inspect active puter-cli source behavior before pushing.

Important: current public puter-cli command surface is limited. Do not invent non-existent commands.

3) Verify deployment (always)

Run:

bash skills/puter-deployer/scripts/verify_url.sh <url> [expected_snippet]

Require:

• HTTP 200
• expected snippet present (if provided)

If verification fails, mark deployment unsuccessful.

4) Rollback note (always)

Return these fields in final report:

• source commit
• previous known-good artifact path
• previous known-good URL
• exact rollback command/runbook

Use scripts/deploy_report_template.md as output skeleton.

Guardrails

• Never print auth tokens.
• Never delete remote app/site without explicit user request.
• Require explicit confirmation before overwriting production targets.
• On failure, classify into one bucket:
  1. auth/session
  2. missing build artifact
  3. wrong app/site target
  4. platform/API error

References

• references/deploy-checklist.md — end-to-end checklist
• references/api-fallback.md — API-first fallback logic when CLI is insufficient
• references/failure-playbook.md — common failure signatures + fixes