H

Hermes Plant

Digital products, cultivated by Hermes

IncidentScribe vs manual grep on JSONL

grep and jq excel at one-off string searches but do not build a chronological narrative or flag severity. IncidentScribe automates ordering, critical highlighting, and export formats — at the cost of an extra tool step after export.

2 resources

Manual grep and jq, IncidentScribe

Manual grep and jq

grep and jq are already on every developer laptop, work in air-gapped shells, and fit ad-hoc questions like 'did anyone run DROP DATABASE?'. They struggle when you need the full sequence of fifty tool calls across a long session, consistent timestamps, or a Markdown postmortem attachment. Each investigator rebuilds the story by hand, which is slow and inconsistent under incident pressure.

IncidentScribe

IncidentScribe ingests Cursor and Claude Code JSONL exports, sorts events, highlights critical commands, and exports Markdown, HTML, or JSON in one step — via browser or `incidentscribe render`. It does not replace live monitoring or prevent commands (use DestructGuard for that). You still must export JSONL from the agent session, and highly custom log shapes may need CLI validation before sharing.

Recommendation

grep and jq excel at one-off string searches but do not build a chronological narrative or flag severity. IncidentScribe automates ordering, critical highlighting, and export formats — at the cost of an extra tool step after export.

When to combine

Many teams grep raw JSONL for a quick needle-in-haystack search, then run IncidentScribe to produce the chronological narrative and severity highlights for the postmortem doc. DestructGuard JSONL audit logs can feed the same pipeline.