← Back to registry
Quality Advanced by Agent Buildprint
Complete Agent Skills Evaluation OS
Prove whether an agent setup is installed, discoverable, useful, safe, reproducible, and worth its context cost.
methodology
agent-ready 01 Contract spine
02 Implementation phases
03 Validation gates
04 Prompt handoff
Executable packet spine
The files or runtime artifacts a fresh agent must read or produce before claiming progress.
- ACTIVATION_EVALS.md
- BUILDPRINT.md
- CONTRACTS.md
- E2E_TASK_BENCH.md
- LOADOUT_INVENTORY.md
- MULTI_AGENT_SAFETY.md
- SAFETY_POLICY.md
- README.md
- SPEC.md
- PLAN.md
- TEST_MATRIX.md
- VALIDATION_TEMPLATE.md
- checks/acceptance.md
Proof gates
Checks that prevent a vague implementation from being reported as complete.
- ✓Offline fixtures only by default
- ✓Scorecard produces evidence
- ✓Process invariants are asserted
- ✓Live adapters require separate acceptance
Risks covered
Failure modes the Buildprint makes visible before an agent can hide them in “done”.
- ⚠Live-provider false positives
- ⚠Undiscoverable skills
- ⚠Unsafe multi-agent execution
What this ships
01
Setup snapshot
02
Static lint
03
Skill unit evals
04
Transcript invariants
05
Scorecard
Validation evidence
For newer phase-flow Buildprints, this names replay or outcome evidence. For older registry entries, it lists the included validation checks.
- Offline fixtures only by default
- Scorecard produces evidence
- Process invariants are asserted
- Live adapters require separate acceptance
Copyable agent prompt
Use the Agent Buildprint: Complete Agent Skills Evaluation OS.
Bootstrap it with `agb start https://agent-buildprint.com/buildprints/complete-agent-skills-evaluation-os/package.json ./my-build` or inspect the GitHub folder. Follow BUILDPRINT.md as the authority spine, read the supporting contracts/checks before implementation, preserve the stated non-goals and safety gates, and do not claim completion until the included validation checks pass.