Assemble the eleven prior lessons into a versioned outputs/agent-workbench-pack/ directory with an installer that lays it idempotently into any target repo.
- Schemas, scripts, and docs from lessons 32 through 40
- The pack layout:
AGENTS.md,docs/,schemas/,scripts/,bin/,README.md,VERSION
outputs/agent-workbench-pack/with the full layout populatedbin/install.sh(orbin/install.py) that refuses to overwrite without--forceVERSIONfile plus aREADME.mddescribing what stays in and what stays out
python3 code/main.pyexits zero and prints the pack tree- Re-running the assembler is idempotent
bin/install.shinto a fresh target leaves a working workbench: state, board, rules, scope, init, runner, gate, reviewer, handoff all in place
- Per-project task content. Tasks belong on the target repo's board, not in the pack.
- Vendor SDK calls. The pack is framework-agnostic by design.
docs/en.md- full lessoncode/main.py- reference implementationoutputs/skill-workbench-pack.md- extracted skill