ai-incidents
Copyright (c) 2026 Holden Salomon

AI disclosure

This tool runs an AI model as part of its normal operation: each run sends
pre-filtered excerpts of your agent transcripts to the model you configure
(the Claude Code CLI by default, or a local model), and that model judges which
moments were real incidents and drafts the ledger entries.

Portions of this repository, including code, tests and documentation, were
written with the assistance of AI (Anthropic Claude / Claude Code) under the
maintainer's direction. The sample ledger and report in examples/ were derived
from entries the tool itself drafted during real use, then curated and
generalized for publication.
