internal/

Package aitools exposes AgentProvenance's read surface, the inline policy gate, and the context-write surface as AI-callable tools.

Package attest provides capture-time, off-host-verifiable signing for evidence - the difference between an audit log and an audit log a defender can trust after host compromise.

Package dashboard serves a local, read-only web view over the verifiable provenance graph: the runs, their merged timeline, the unified signals/risks, the verify+signature status, and -- the signature view -- the causality DAG (LLM intent -> action -> policy -> risk), which a flat event stream cannot show.

Package hooksbridge translates a Claude Code (or compatible) harness hook log into AgentProvenance's graph: agent nodes, delegation spawn edges, peer (peer) message edges carrying the message body as objectified evidence, and a tool_call per agent tool invocation bound to the acting agent_id.

Package intent turns the model/agent's declared intent and the system's observed runtime effects into a first-class Intent-Runtime Diff.

Package launch is the one-command porcelain entry point for AgentProvenance.

Package mcpserver exposes AgentProvenance's AI tool surface over the Model Context Protocol (MCP, spec revision 2025-06-18) as a stdio JSON-RPC 2.0 server.

Package producer models evidence-producer deployment: where a sensor runs, how scope is resolved, how events ship, and which collection layers a given environment can actually deliver.

Package redact masks secrets out of captured evidence before it is stored, materialized into content-addressed objects, or exported into a signed forensics bundle.

Package signals is the unified, graph-attached signal model - the single row type that every observability dimension lands on.

Package tlsintent reassembles the fragmented TLS plaintext that the eBPF SSL_write/SSL_read probes capture into COMPLETE HTTP/1.1 request/response messages, and pulls minimal LLM semantics (model, endpoint) out of the body.