Log Intake — a Spec Kit extension for logreduce

Adds /speckit.logreduce.intake, a log-intake step for Spec Kit's bundled bug triage workflow. It runs logreduce on the logs attached to a bug report and saves the result as .specify/bugs/<slug>/evidence.md. /speckit.bug.assess, fix and test then work from one saved, reviewable, reduced log (8k tokens by default) instead of each re-reading a raw log that may run to thousands or millions of lines.

Install

1. Put the logreduce binary (≥ 0.2.0) on your PATH — see the releases page.

2. Add the bundled bug workflow and this extension to your Spec Kit project:

   specify extension add bug
   specify extension add logreduce --from \
     https://github.com/alexcpn/log_tfidf_reducer/releases/latest/download/logreduce-speckit.zip
   

   To install from a checkout: specify extension add --dev path/to/log_tfidf_reducer/speckit.

Use

/speckit.logreduce.intake slug=deploy-timeout ci/run-4812.log
/speckit.bug.assess slug=deploy-timeout Deploy step times out; log in ci/run-4812.log, evidence in .specify/bugs/deploy-timeout/evidence.md

Options in the input: slug=<name> and budget=<tokens> (default 8000). Running intake again (for example after /speckit.bug.test fails) adds a new dated section to evidence.md rather than overwriting the earlier one.

As a hook

The extension registers an optional before_bug_assess hook, so /speckit.bug.assess offers the intake step itself — once the bundled bug extension emits that event. That change is proposed upstream. Until it lands, run intake yourself first, as shown above.

What goes in evidence.md

For each log: the source path, the original line count, the exact logreduce command, its stats, and the reduced log verbatim. Logs under ~500 lines are listed as "read directly". Secret/PII redaction stays on.