task-autoresearch-report

Task Autoresearch — Report

Render the final task-autoresearch report using the shape from templates/REPORT.md (in this plugin):

## Summary

## Confirmed

## Inferred

## Evidence

## Changes

## Verification

## Remaining Risk

Rules

• Pull content from the current conversation's evidence trail — commands run, files inspected, artifacts extracted, verification results.
• Confirmed is for facts you observed directly. Inferred is for conclusions drawn from those facts. Never blend them.
• Evidence lists concrete pointers: file paths with path:line, command outputs (truncated, with the truncation marked), artifact filenames. Not narrative.
• Changes lists changed files and a one-line "what changed" each — no diffs unless the user asked for them.
• Verification lists the exact commands run and their pass/fail outcome.
• Remaining Risk lists what is still unknown or untested. If nothing remains, say "None known" — do not omit the section unless the task is genuinely tiny.
• Omit other sections only if they are truly empty.

Output

If the user passed a target path, write the report there. Otherwise write to ./REPORT.md. If ./REPORT.md already exists, append a numeric suffix rather than overwriting.

Keep the whole report under ~200 lines unless the user asked for a full dossier.