# Complete Agent Skills Evaluation OS - Agent Guide

Complete evaluation stack for agent skill setups: validate config, measure loadout cost, test skills, check activation, enforce transcript process, and score multi-agent safety.

## Agent instruction

Do not scrape the human UI. Use this agent guide, the package manifest, and raw Buildprint files.

1. Fetch package manifest: `/buildprints/complete-agent-skills-evaluation-os/package.json`
2. Read order: `BUILDPRINT.md`.
3. Do not scrape human cards. Use this manifest, agent.md, and raw files. BUILDPRINT.md is the canonical start file and owns the required read order, phase gates, and acceptance gates. Structured control files are machine-readable mirrors only.
4. Follow the Buildprint's alignment/question rules before implementation.
5. Run required validation and write requested validation evidence plus the final chat handover.

## Metadata

- slug: `complete-agent-skills-evaluation-os`
- category: `Workflow OS`
- tier: `agent-grade`
- status: `validated`
- runtime: Any coding agent, JavaScript proof, CI-ready adapters
- stack: Static lint, Loadout inventory, Skill tests, Activation evals, Transcript checks, Multi-agent safety

## Entrypoints

- Human page: /buildprints/complete-agent-skills-evaluation-os/
- Manifest JSON: /buildprints/complete-agent-skills-evaluation-os/package.json
- Prompt: /buildprints/complete-agent-skills-evaluation-os/prompt.txt
- GitHub: https://github.com/DomEscobar/agent-buildprint/tree/main/buildprints/complete-agent-skills-evaluation-os
- Raw base: https://agent-buildprint.com/buildprints/complete-agent-skills-evaluation-os/files

## Files

- `ACTIVATION_EVALS.md` - Buildprint package file (required)
- `BUILDPRINT.md` - compatibility bootstrap or package contract (required)
- `checks/acceptance.md` - acceptance checklist (required)
- `CONTRACTS.md` - legacy interface/data contracts, when present (required)
- `E2E_TASK_BENCH.md` - Buildprint package file (required)
- `LOADOUT_INVENTORY.md` - Buildprint package file (required)
- `MULTI_AGENT_SAFETY.md` - Buildprint package file (required)
- `PLAN.md` - legacy execution index, when present (required)
- `proof/package-lock.json` - offline proof artifact (required)
- `proof/package.json` - offline proof artifact (required)
- `proof/src/eval-os.mjs` - offline proof artifact (required)
- `proof/test/eval-os.test.mjs` - offline proof artifact (required)
- `publication.json` - machine-readable mirror (required)
- `README.md` - human overview, non-authoritative (required)
- `SAFETY_POLICY.md` - Buildprint package file (required)
- `SCORECARD.md` - Buildprint package file (required)
- `SKILL_UNIT_EVALS.md` - Buildprint package file (required)
- `SPEC.md` - legacy behavior requirements, when present (required)
- `STATIC_LINT.md` - Buildprint package file (required)
- `TEST_MATRIX.md` - legacy risk-to-test alignment, when present (required)
- `TRANSCRIPT_PROCESS_EVALS.md` - Buildprint package file (required)
- `VALIDATION_REPORT.md` - Buildprint package file (required)
- `VALIDATION_TEMPLATE.md` - legacy completion report template, when present (required)

## Copyable implementation prompt

```txt
Use the Complete Agent Skills Evaluation OS Buildprint. First bootstrap exact snapshots: agb start https://agent-buildprint.com/buildprints/complete-agent-skills-evaluation-os/package.json . If agb is not installed, clone https://github.com/DomEscobar/agent-buildprint and run node agent-buildprint/bin/agb.js start https://agent-buildprint.com/buildprints/complete-agent-skills-evaluation-os/package.json . Then read .buildprint/next-agent.md and continue. Do not write Buildprint snapshots manually. Build the evaluation stack from the packet; use offline fixtures by default and ask before live external actions.
```