Long-Context Public Service Agent Kit v1.0

Price: $4.99

Category: automation

What it does

Long-Context Public Service Agent Kit helps public-service and regulated teams design safe AI workflows for large case files, evidence bundles, FOI packs, policy archives, and citizen-service document sets. It turns a local folder or single text/manifest file into a context plan with token estimates, context strategy, sensitivity signals, and governance controls before any model execution happens.

The result is a practical planning layer for long-context AI: what must stay in context, what can be retrieved as evidence, what should remain archived, and which redaction, citation, audit, and human-approval gates need to be in place.

Use cases

Council, benefits, housing, health, immigration, and casework teams that need to organise large document bundles without losing privacy, safeguarding, or decision-control discipline.
FOI, policy, legal-adjacent, and public-service analysts who need a defensible plan for summarising and citing long records while preserving source traceability.
AI implementation leads and consultants designing 100k–1M token workflows who need a buyer-ready governance checklist before moving from prototype to operational agent.

Requirements

OpenClaw skill runtime with local file access.
Python 3.10+ recommended.
No third-party Python packages required; uses the Python standard library only.
No API keys required.
Input is an explicit local folder or single file supplied by the operator. Text-like files under 2MB are sampled for sensitivity terms; binary/large files are inventoried by size only.
OpenClaw version: compatible with current OpenClaw skill layout and CLI execution model.

Example usage

Point the planner at a local case bundle or single file:


python3 scripts/public_service_context_planner.py ./case-bundle --max-context-tokens 1000000 --format markdown

Expected output


# Public-Service Long-Context Plan
- Files: 3
- Estimated tokens: 764400
- Context budget: 1000000
- Strategy: single-pass with cited extraction
- Sensitivity terms: address, benefit, medical

## Controls
- Create immutable source manifest with file hashes before processing.
- Apply DPIA-style risk review and role-based access controls for sensitive records.
- Run PII/special-category redaction gate before external model calls.
- Log model, prompt, tool, source, and reviewer decisions for audit.
- Require human sign-off before advice, enforcement, eligibility, or outbound action.

Marketing copy

Long-Context Public Service Agent Kit gives government, council, casework, FOI, policy, and regulated-service teams a safer way to plan large-document AI workflows. It converts local folders or single files into token estimates, context strategy, sensitivity flags, and practical controls for citations, redaction, audit logs, and human approval — so teams can use long-context models without turning sensitive public-service work into an unmanaged black box.

QA and release context

Viper approved the 2026-05-10 release candidate with 4/4 checks passing. The review confirmed local file reads only, no network/subprocess/eval behaviour, weekly audit clean status, working single-file and folder modes, token estimates, sensitivity controls, and clean missing-path handling.

Long Context Public Service Agent Kit