Skip to content

docs: add review-learnings doc distilled from recent PRs#3057

Open
vseehausen wants to merge 2 commits into
mainfrom
valentin/pr-review-learnings-skill
Open

docs: add review-learnings doc distilled from recent PRs#3057
vseehausen wants to merge 2 commits into
mainfrom
valentin/pr-review-learnings-skill

Conversation

@vseehausen
Copy link
Copy Markdown
Member

@vseehausen vseehausen commented Apr 24, 2026

Summary

Adds docs/review-learnings.md with review conventions distilled from recent PRs, and a pointer to it in AGENTS.md.

Changes

  • New docs/review-learnings.md — conventions grouped by topic: module layout, naming, types & schemas, reuse, controllers, auth, DB, logging, config, testing, code style, security, PR hygiene, Biome.
  • AGENTS.md — 4-line pointer to the new file.

Ledger line at the top of the new file records the analyzed window so future runs of /safe-engineering:pr-learnings can resume from there.

@vseehausen vseehausen requested a review from a team as a code owner April 24, 2026 15:00
@vseehausen vseehausen changed the title docs: split review learnings into dedicated file docs: add review-learnings doc distilled from recent PRs Apr 24, 2026
Copy link
Copy Markdown
Contributor

@0xAryanShah 0xAryanShah left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Nice!!

@vseehausen vseehausen force-pushed the valentin/pr-review-learnings-skill branch from 8424a30 to bc489ff Compare April 27, 2026 08:34
Move the growing set of PR review conventions out of AGENTS.md into
docs/review-learnings.md, leaving a one-paragraph pointer behind.

Why: AGENTS.md auto-loads into every Claude Code session in the repo, so
every token there costs context across all work. Review learnings grow
over time as we extract more patterns from PR comments; keeping them in
a dedicated file lets the conventions compound without inflating the
always-on context.

The separate file is maintained semi-automatically by the new
/safe-engineering:pr-learnings skill (opened as a separate PR on
safe-global/safe-engineering-plugin), which scans recent PR comments on
each run, extracts recurring patterns, merges them into existing rules
without duplicating, and updates a resume ledger at the top of the file.

Initial content distilled from 412 human review comments across 74 PRs
in the 2026-02-23 → 2026-04-23 window. Ledger line at the top of the
file records the range and the date to resume from next run.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
@vseehausen vseehausen force-pushed the valentin/pr-review-learnings-skill branch from bc489ff to 750a838 Compare April 27, 2026 08:41
@vseehausen vseehausen force-pushed the valentin/pr-review-learnings-skill branch from 0386224 to 8131194 Compare April 27, 2026 09:37
@vseehausen vseehausen enabled auto-merge (squash) April 28, 2026 13:05
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.

2 participants