Establish Forgejo compatibility and review-quality gates #26

New issue

Open

opened 2026-07-14 22:32:30 -04:00 by caleb-brown · 0 comments

caleb-brown commented

2026-07-14 22:32:30 -04:00

Owner

What to build

Maintainers can prove that Temper supports its declared Forgejo envelope and produces acceptably grounded, actionable review feedback before a release is approved.

Acceptance criteria

A versioned Forgejo contract suite covers every relied-upon webhook, permission, pagination, diff, status, comment, review, inline-position, rate-limit, and token-scope behavior.
The supported Forgejo version envelope and rejection behavior are documented from the contract evidence.
A versioned representative PR corpus covers multiple languages, review profiles, no-findings cases, large changes, and adversarial content.
A human rubric scores correctness, actionability, novelty, severity, diff grounding, duplicate suppression, coverage, and harmful noise.
Security acceptance covers prompt injection, hostile rendering, fabricated locations, unauthorized forks, oversized input, and credential leakage.
Release thresholds and the process for reviewing corpus regressions are explicit and reproducible.
The gate fails clearly when compatibility, security, or quality evidence falls below the accepted bar.

Blocked by

#17 — Publish safe inline findings
#19 — Add manual, skip, and fork-review controls
#21 — Manage Review Policies and Reviewer Profiles safely

## What to build Maintainers can prove that Temper supports its declared Forgejo envelope and produces acceptably grounded, actionable review feedback before a release is approved. ## Acceptance criteria - [ ] A versioned Forgejo contract suite covers every relied-upon webhook, permission, pagination, diff, status, comment, review, inline-position, rate-limit, and token-scope behavior. - [ ] The supported Forgejo version envelope and rejection behavior are documented from the contract evidence. - [ ] A versioned representative PR corpus covers multiple languages, review profiles, no-findings cases, large changes, and adversarial content. - [ ] A human rubric scores correctness, actionability, novelty, severity, diff grounding, duplicate suppression, coverage, and harmful noise. - [ ] Security acceptance covers prompt injection, hostile rendering, fabricated locations, unauthorized forks, oversized input, and credential leakage. - [ ] Release thresholds and the process for reviewing corpus regressions are explicit and reproducible. - [ ] The gate fails clearly when compatibility, security, or quality evidence falls below the accepted bar. ## Blocked by - [#17](https://git.caleb-brown.dev/caleb-brown/temper/issues/17) — Publish safe inline findings - [#19](https://git.caleb-brown.dev/caleb-brown/temper/issues/19) — Add manual, skip, and fork-review controls - [#21](https://git.caleb-brown.dev/caleb-brown/temper/issues/21) — Manage Review Policies and Reviewer Profiles safely

caleb-brown added the

ready-for-agent

label

2026-07-14 22:32:30 -04:00

caleb-brown referenced this issue

2026-07-14 22:32:32 -04:00

Prove the complete v1 operating envelope #28

caleb-brown added a new dependency

2026-07-14 22:38:29 -04:00

#17 Publish safe inline findings

caleb-brown added a new dependency

2026-07-14 22:38:30 -04:00