reverse-sprinter

Reverse Sprinter — Delegate a Reverse Sprint

Use this skill to delegate a reverse sprint to a low-cost developer agent (a developer row from harnez agent models). You are the delegator, not the executor.

1. Capture the task context exactly as given — ticket numbers, scoped task or bug text, target files, and acceptance criteria.

2. Select a low-cost coder model (the cheapest fitting developer row of harnez agent models, or a user-specified model).

3. Spawn exactly one subagent on that low model with this instruction:

   Read the installed reverse-sprint skill completely before doing anything else. You are the Lead Developer driving this task. Execute that skill's complete reverse-sprint workflow yourself: implement the code, author tests, perform frequent auto-compaction (every 100-150k tokens), request milestone reviews from a reviewer model of another vendor, and escalate to an advisor model (both chosen from harnez agent models) with strict limited context only if stuck. Report back when teardown and tracker sync are complete.

   Task: <verbatim task context>

4. Do not restate or summarize the reverse workflow — the reverse-sprint skill is the single source of that policy.

5. Stay responsive after dispatch. Report the handoff and subagent ID to the user.