Find past investigations whose alert text is similar to the current alert, ranked by embedding cosine similarity, with their recorded root cause and a link back to the original incident thread.
Installs into .claude/skills of the current project.
Are you the author of Find Similar Incidents?
Add the live security badge to your README. It updates with every re-scan.
[](https://www.skillsdirectory.com/skills/perun-engineering-find-similar-incidents)
---
name: find_similar_incidents
description: Find past investigations whose alert text is similar to the current alert, ranked by embedding cosine similarity, with their recorded root cause and a link back to the original incident thread.
tool: agents.incident_history.tools:find_similar_incidents
---
# When to use
Call this skill once per investigation, when the user message is a serialized `AlertContext` (the default investigation request). Pass the `alert_text` field from the alert verbatim.
# Inputs
- `alert_text` (required): the full text of the current alert. The tool embeds it with Amazon Titan and ranks stored incident-outcome records by cosine similarity.
# Output
A human-readable summary plus an embedded `AGENT_RESULT` footer. Each finding is one similar past incident:
- `Similar alert (~NN% match) fired YYYY-MM-DD: "<alert preview>". Root cause: <recorded root cause>. <summary>`
- The finding's link points back to the original incident thread when one was recorded.
No matches is a normal, successful outcome — the tool returns "No similar past incidents found." Do not invent incidents; relay only what the tool ranked.
