Skip to content

feat(review): consolidate output into one numbered report with best-practices checks#210

Merged
ryzizub merged 5 commits into
mainfrom
claude/kind-fermi-092663
Jul 3, 2026
Merged

feat(review): consolidate output into one numbered report with best-practices checks#210
ryzizub merged 5 commits into
mainfrom
claude/kind-fermi-092663

Conversation

@ryzizub

@ryzizub ryzizub commented Jul 2, 2026

Copy link
Copy Markdown
Contributor

Description

Overhauls the review output into one numbered report and hardens the review skill's authoring.

  • One consolidated report (feat(review): consolidate output into a single file #180)/review, /build, and /hotfix now emit a single review file instead of one per agent. Per-agent detail moves to a linked raw/ subdir; /review isolates each run under docs/code-review/<slug>/; ephemeral cleanup is unchanged.
  • Stable, numbered, aligned findings (fix(review): align chat summary with review content #179, feat: number review findings #194) — agents return structured findings; the skill deduplicates, orders deterministically, and assigns stable FINDING-NN ids plus a <category>/<rule> class id. The chat summary and the file render from the same list (same ids, order, titles), so you can act by number ("fix all except 3 and 4") or by rule.
  • Authoring fixes — one consistent RAW_DIR/<name> contract, one chat-summary rule, a defined Critical/Important/Suggestion taxonomy with agent mapping, a deterministic scope slug, and each skill owning its own post-review menu.

Verified with an adversarial review workflow (15/16 defects resolved on the first pass, remainder fixed after); markdownlint, cspell, and claude plugin validate all clean.

Closes #179
Closes #180
Closes #194

Type of Change

  • New feature (feat)
  • Bug fix (fix)
  • Code refactor (refactor)
  • Documentation (docs)
  • CI change (ci)
  • Chore (chore)

🤖 Generated with Claude Code

ryzizub and others added 2 commits July 2, 2026 14:04
…ractices checks

Emit a single consolidated review file per run instead of one file per agent.
Each agent writes its detailed report to a raw/ subdirectory and returns a
structured findings list; the calling skill deduplicates, orders deterministically
(severity → file → line → rule), and assigns stable FINDING-NN ids plus a
<category>/<rule> class id. The chat summary renders from that same ordered list,
so ids, order, and titles match between chat and file, and the user can act by
finding id or by rule. Applies to /review, /build, and /hotfix; ephemeral cleanup
still removes the whole output directory.

Add best-practices-review-agent: validates changed code against current official
documentation and best practices (deprecation checks, citations), including a
Claude/Anthropic SDK checklist gated to code that uses the SDK. Wired into /review
and /build, offered conditionally in /hotfix.

Closes #179
Closes #180

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
ryzizub and others added 2 commits July 3, 2026 11:40
…defects

Drop the best-practices-review-agent and its wiring — the "review against docs"
reading was a misunderstanding; the ask was to author the review skill well.
Build returns to 5 agents, /review to 4, hotfix to its reduced set.

Fix the cross-file contract defects found auditing the skill against Anthropic
Claude Code authoring standards:
- `<name>` is a bare stem so agents write `<RAW_DIR>/<name>.md` (no raw/raw/x.md.md)
- build/hotfix resolve an absolute RAW_DIR via pwd
- the shared "Acting on findings" step is neutral; each skill owns its post-review menu
- one chat-summary contract: Critical+Important verbatim, Suggestions collapsed
- define the Critical/Important/Suggestion taxonomy and how agents map onto it
- drop location "-"; define where a bare-file finding sorts
- /review isolates each run under docs/code-review/<slug>/ (no raw/ clobber)
- correct the id-stability claim; specify the merge survivor rule id; custom-agent derivation
- deterministic path-scope slug; build/hotfix commits don't cite deleted-report ids
- hotfix gains agent-failure handling; correct the "raw not read back" claim

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
# Conflicts:
#	config/cspell.json
#	skills/build/SKILL.md
@ryzizub ryzizub marked this pull request as ready for review July 3, 2026 10:26
@ryzizub ryzizub requested a review from a team as a code owner July 3, 2026 10:26
@ryzizub ryzizub merged commit 47774b2 into main Jul 3, 2026
7 checks passed
@ryzizub ryzizub deleted the claude/kind-fermi-092663 branch July 3, 2026 14:00
@vgvbot vgvbot mentioned this pull request Jul 3, 2026
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.

feat: number review findings feat(review): consolidate output into a single file fix(review): align chat summary with review content

2 participants