PSF evidence pack

Generate the files a production AI repository should expose.

Create a ready-to-commit evidence pack for the Agent Readiness Index. It gives maintainers an immediate path to publish evals, output validation, incident response, autonomy gates, and provider fallback evidence without waiting for PAI review.

Project nameOrganisationRepository URLSystem typeAutonomy levelEvidence ownerContact

README.mdREADME.md

production-ai-readiness.mddocs/production-ai-readiness.md

autonomy-matrix.mddocs/autonomy-matrix.md

output-validation.mddocs/output-validation.md

README.mddocs/evals/README.md

incident-response.mddocs/incident-response.md

provider-fallbacks.mddocs/provider-fallbacks.md

production-ai-readiness.md.github/ISSUE_TEMPLATE/production-ai-readiness.md

Preview

README.md

The ZIP also includes docs for evals, incidents, autonomy, output validation, and fallback evidence.

# Customer support agent - Production AI Evidence Pack

This repository publishes visible evidence for a production AI system using the Production AI Institute Production Safety Framework (PSF).

This evidence pack is not certification. It is a structured way to make readiness evidence inspectable by maintainers, customers, auditors, and reviewers.

## System Profile

- Organisation: Example AI Team
- System type: Customer-facing agent
- Autonomy level: Drafts for human review
- Evidence owner: AI platform lead
- Contact: ai-governance@example.com
- Repository: https://github.com/org/repo
- Last reviewed: 2026-06-24

## PSF Evidence Index

| Domain | Evidence file | Status |
| --- | --- | --- |
| D1 Input boundary | docs/production-ai-readiness.md | Draft |
| D2 Output validation | docs/production-ai-readiness.md | Draft |
| D3 Data stewardship | docs/production-ai-readiness.md | Draft |
| D4 Observability | docs/production-ai-readiness.md | Draft |
| D5 Deployment control | docs/production-ai-readiness.md | Draft |
| D6 Human oversight | docs/production-ai-readiness.md | Draft |
| D7 Security posture | docs/production-ai-readiness.md | Draft |
| D8 Ecosystem resilience | docs/production-ai-readiness.md | Draft |

## Minimum Evidence To Add

- Scope and allowed inputs.
- Output schema or validation contract.
- Data handling and retention notes.
- Eval set, scoring method, and review cadence.
- Release gate and rollback notes.
- Human approval and escalation rules.
- Security and tool-permission notes.
- Provider fallback or degraded-mode notes.

## Public Report

Run the PAI Agent Readiness Index:

https://www.productionai.institute/agent-readiness