architect: bug-report pipeline — inbound classification + auto-close #24

Merged

disinto-admin merged 1 commit from architect/bug-report-pipeline into main

2026-04-15 17:38:55 +00:00

architect-bot commented

2026-04-12 04:10:10 +00:00

Collaborator

What this sprint enables

After this sprint, bug-reports flow through a cheap classification gate before reaching the expensive reproduce-agent. Inspection-class bugs (stack trace cited, cause obvious from code) go straight to dev-agent — saving the full Playwright/MCP environment spin-up. The auto-close loop fires reliably, and upstream Codeberg reporters get notified when their bug is fixed.

Today: every bug-report → reproduce-agent (expensive). After: only ambiguous bugs → reproduce-agent; obvious bugs → dev-agent directly.

What exists today

The pipeline is 80% built: gardener detection + enrichment, reproduce-agent with Playwright, triage-agent 6-step workflow, dev-poll label gating, auto-close logic (not firing), issue templates, manifest action system.

Complexity

5-6 sub-issues estimated
~8 files touched across formulas, lib, and gardener
Mostly gluecode — extending existing gardener formula, adding classification step, wiring auto-close reliability
No new containers or services

Risks

Classification accuracy (conservative criteria mitigate)
Gardener formula complexity (separate step mitigates)
Upstream Codeberg notification needs vault action (AD-006)

Cost — new infra to maintain

One new gardener formula step (inbound classification)
2-4 bug taxonomy labels
One vault action template for upstream notification
No new services, cron jobs, or agent roles

Recommendation

Worth it. Infrastructure is 80% built. Fills two concrete gaps (classification gate + auto-close reliability) with minimal new maintenance. Biggest value: avoiding unnecessary reproduce-agent runs.

Reply ACCEPT to proceed with design questions, or REJECT: <reason> to decline.

## What this sprint enables After this sprint, bug-reports flow through a **cheap classification gate** before reaching the expensive reproduce-agent. Inspection-class bugs (stack trace cited, cause obvious from code) go straight to dev-agent — saving the full Playwright/MCP environment spin-up. The auto-close loop fires reliably, and upstream Codeberg reporters get notified when their bug is fixed. Today: every bug-report → reproduce-agent (expensive). After: only ambiguous bugs → reproduce-agent; obvious bugs → dev-agent directly. ## What exists today The pipeline is 80% built: gardener detection + enrichment, reproduce-agent with Playwright, triage-agent 6-step workflow, dev-poll label gating, auto-close logic (not firing), issue templates, manifest action system. ## Complexity - **5-6 sub-issues** estimated - **~8 files** touched across formulas, lib, and gardener - **Mostly gluecode** — extending existing gardener formula, adding classification step, wiring auto-close reliability - **No new containers or services** ## Risks - Classification accuracy (conservative criteria mitigate) - Gardener formula complexity (separate step mitigates) - Upstream Codeberg notification needs vault action (AD-006) ## Cost — new infra to maintain - One new gardener formula step (inbound classification) - 2-4 bug taxonomy labels - One vault action template for upstream notification - No new services, cron jobs, or agent roles ## Recommendation **Worth it.** Infrastructure is 80% built. Fills two concrete gaps (classification gate + auto-close reliability) with minimal new maintenance. Biggest value: avoiding unnecessary reproduce-agent runs. --- Reply `ACCEPT` to proceed with design questions, or `REJECT: <reason>` to decline.

architect-bot added 1 commit 2026-04-12 04:10:10 +00:00