Skip to content

Add aria-alert-investigator skill for Aria health-metric IcMs, Fixes AB#3589793#428

Open
somalaya wants to merge 2 commits intomasterfrom
skills/aria-alert-investigator
Open

Add aria-alert-investigator skill for Aria health-metric IcMs, Fixes AB#3589793#428
somalaya wants to merge 2 commits intomasterfrom
skills/aria-alert-investigator

Conversation

@somalaya
Copy link
Copy Markdown
Contributor

@somalaya somalaya commented Apr 30, 2026

Adds a new skill at .github/skills/aria-alert-investigator/SKILL.md that handles Aria-detected anomaly IcMs (titles starting with Aria detected an incident in <project> for <metric>).

Why

These alerts have a fundamentally different investigation pattern than customer-reported incidents: there is no customer, no error chain, and no log file — only "the curve moved". The existing incident-investigator skill assumes customer-reported failures and produces wrong conclusions on Aria alerts. I have hit this several times now while triaging anomaly-detection IcMs.

What the new skill enforces

  • Decode the metric → KQL slice (BLOCKING) — Aria health-metric names are marketing-style labels; never run queries until the slice is confirmed (or assumption is explicitly flagged).
  • Treat the alert value as opaque — don't reverse-engineer "4.48 in Red band" into "4 events".
  • Multi-angle trend analysis — raw count, normalized rate, traffic volume, same-day-of-week, hourly-around-alert. A single view fools you.
  • Past ICMs are pattern signal, not verdict — call out the family pattern, but don't conclude with it.
  • Code context is the baseline, not extra step — delegate continuously to codebase-researcher to understand what unfamiliar attributes/error codes mean.
  • Strict output format — data findings stated first; opinions only in a separate "My read" section with explicit confidence and basis.

Other changes

  • Register the skill in .github/copilot-instructions.md so the agent picks it up.
  • Add a pointer from kusto-analyst SKILL.md so Kusto-query authors delegate the Aria-specific workflow back here.
  • Fix two unrelated pre-existing typos in kusto-analyst/SKILL.md (corrupted MATS table row +------=+| and isspecified missing space).

Tested via

Used the new workflow end-to-end on IcM 786276130 (TD - ATS Devices W/ Success Aria alert) earlier today. Multi-angle trend correctly attributed the metric drop to the 14.2.0 broker rollout instead of the alert grain itself, matching the historical pattern from sibling ICMs.

Fixes AB#3589793

@somalaya somalaya requested a review from a team as a code owner April 30, 2026 04:15
@github-actions
Copy link
Copy Markdown

❌ Work item link check failed. Description does not contain AB#{ID}.

Click here to Learn more.

Adds a new skill at .github/skills/aria-alert-investigator/SKILL.md that handles Aria-detected anomaly IcMs (titles starting with 'Aria detected an incident in <project> for <metric>'). These alerts have a fundamentally different investigation pattern than customer-reported incidents: there is no customer, no error chain, and no log file, only 'the curve moved'. The new skill enforces evidence-first investigation with metric-slice confirmation, multi-angle trend analysis, and explicit separation of data findings from opinions.

Also:

- Register the skill in .github/copilot-instructions.md so the agent picks it up.

- Add a pointer from the kusto-analyst skill so authors of Kusto queries delegate the Aria-specific workflow back here.

- Fix two unrelated pre-existing typos in kusto-analyst SKILL.md (corrupted MATS table row and 'isspecified' missing space) that would otherwise show up in the same diff.
@somalaya somalaya force-pushed the skills/aria-alert-investigator branch from 25b34aa to 5b32dbf Compare April 30, 2026 04:18
@somalaya somalaya changed the title Add aria-alert-investigator skill for Aria health-metric IcMs Add aria-alert-investigator skill for Aria health-metric IcMs, Fixes AB#3589793 Apr 30, 2026
@github-actions
Copy link
Copy Markdown

✅ Work item link check complete. Description contains link AB#3589793 to an Azure Boards work item.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant