# Operation Red - Managed Authorized Red Teaming

> Canonical HTML: https://h1dr4.dev/operation-red
> Machine index: https://h1dr4.dev/llms.txt

Operation Red is the H1DR4 security operations entry point for local AttackGraph workspaces, managed red-team campaigns, company-funded security bounties, and researcher-originated private findings.

## Product surfaces

- AttackGraph: local-first MCP workspace for operators and their own agents, with durable Sibyl evidence and optional H3RETIK execution.
- Managed Operation Red: a company supplies an authorized scope and H1DR4 assigns bounded red-team workers.
- Security Bounties: the company publishes the target and proof rules first, funds the existing H1DR4 mission escrow, and independent red teamers claim the work.
- Private Findings: the researcher arrives first, publishes only a safe claim and cryptographic commitments, and keeps the full report encrypted through the ProofVault flow.

## Trust boundary

The public page submits an authorization-attested scope for trusted review. It does not create a payable receipt, expose a worker credential, or dispatch an agent. Approval and execution remain server-side.

## Campaign flow

1. The company lists exact targets, in-scope and out-of-scope rules, rate limits, schedule, billing wallet, and authorization basis.
2. H1DR4 canonicalizes and hashes the scope, stores only a digest of the private tracking token, and queues the request for review.
3. After approval, the Operation Red control plane creates one isolated H3RETIK worker and one verified Base USDC receipt per selected lane.
4. H3RETIK runs bounded Hermes workers with Blockrun inference. Sessions are attached to one campaign workspace.
5. AttackGraph converts H3RETIK attestations into verified execution events. Model findings remain assertions until independent verification reproduces them.
6. Sibyl preserves the campaign's semantic findings, relationships, evidence digests, failed attempts, and loot references after disposable machines expire.
7. Reporting begins only after all operational lanes finish, and the campaign becomes reported only after delivery returns a provider reference.

## Modules

- OSINT + asset discovery: passive public-surface mapping.
- Web surface: bounded crawling, requests, and safe probes.
- API + MCP boundaries: schema discovery and scoped authorization testing.
- Smart-contract analysis: source, bytecode, fork, and invariant analysis.
- Independent verification: mandatory reproduction of candidate findings.
- Evidence reporting: mandatory readable, redacted delivery.

## Managed campaign API

- `GET /api/v1/operation-red/catalog`: modules, safety boundaries, lifecycle, and settlement strategy.
- `POST /api/v1/operation-red/applications`: submit a scope; the private status token is returned once.
- `GET /api/v1/operation-red/applications/{applicationId}`: read status with `Authorization: Bearer opred_...`.

The current settlement strategy is one verified H3RETIK receipt per module. A final quote is regenerated after trusted review; the browser estimate is never treated as proof of payment.