Skip to content

lyddonb/bstack

v0.13.1MIT

Beau's rigorous agent workflow. beau-mode routes each task to a playbook and the principles; beau-help lists what the kit can do and what's new; skills for understanding (how, why, teach, recall), debugging (root-cause finds a bug's cause from facts, not from reading the code), proof (verify-work, create- and maintain-verification-skill, create- and maintain-formal-verification-skill, swarm), design and planning (architect, arena, state-machines, technical-writing), product repos (scaffold-product and the Product phase playbook), review and trails (interrogate, show-me-your-work), improvement (reflect, correct, unslop), and test economics (qualitydirector records per-test time and outcome, books the runs into the quality books, checks them and summarizes a week) do the steps. Adapted from Lauren Tan's pstack to work in any agent.

root-cause

Find a bug's root cause from facts instead of from reading the code. Keeps a case file of facts, hypotheses and the matrix between them, steers by the strongest signal, has an adversarial reviewer on another model check the work, and refuses to name a root cause until exactly one hypothesis explains every fact. Use for "root cause this", "find out why this broke", "debug without reading the code", a bug or regression in agent-written or unfamiliar code, or after two theories have failed. Also turns a solved case into a regression or deploy gate with recheck.

Version
0.1.0
License
MIT
Read SKILL.md at the source

Pinned to revision b91bbd04e882, so it is the text this page describes rather than whatever the author pushed since.

Files

Every link opens the file at its source, pinned to the revision this page describes.