borglisted

# Borg — Failure Memory for AI Coding Agents Borg helps agents check prior rescue guidance before repeating known debugging dead ends. When there is no confident match, it should return `NO_CONFIDENT_MATCH` instead of forcing advice. ## What is Borg? Borg maintains a registry of **workflow packs** — structured approaches to debugging, testing, code review, planning, deployment, and more. Each pack should expose its confidence level, known failure cases, checkpoints, and verification rules. Think of it as: *your agent's peer support group*. Before going in circles for the fourth attempt, ask borg. ## When to Use This Skill Apply this skill when ANY of the following are true: - **Stuck in a loop** — 3+ failed attempts on the same problem - **Debugging burnout** — spending significant token budget with no clear root cause - **Need structure** — task would benefit from a proven phase-by-phase approach - **No relevant context** — working in an unfamiliar domain without a mental model - **Test failures** — unclear how to approach recovery after test failures - **Code review confusion** — unsure what to look for or how to prioritize findings ## How to Use ### Step 0: Concrete Failure Rescue First If you have an exact error, failing command output, traceback, or install/config/deploy failure, do **not** start with pack search. Ask Borg for the day-one rescue packet first: ```text Tool: error_lookup Args: input="<exact error or failing command output>", show_guidance=false