flow-next-impl-review

github:gmickel/gmickel-claude-marketplace · first committed 2026-01-06 · source · badge

John Carmack-level implementation review via RepoPrompt or Codex. Use when reviewing code changes, PRs, or implementations. Triggers on /flow-next:impl-review.

Skill Health: not yet scored

Listed, not yet scored. This skill has no public model test yet. It is in the scoring queue; submitted skills go first, then the most-installed.

Version history

92 commits to this file on GitHub (newest first).

2026-10-31 · commit 3796bc0 · Decided audit changes, each with its test (audit 2026-10 wave 3) (#588)
2026-10-10 · commit d5660ce · feat(review): task reviews carry a review focus; Windows jq fix; release 8.5.0 (#540)
2026-10-05 · commit 03ae8e1 · fix(review): review failures name their cause, a rewritten spec loses its plan SHIP, land reports a queued merge
2026-10-04 · commit c5b2b00 · fix(review): fixes from the first review round
2026-10-04 · commit 6102845 · feat(review): one review runner with the same panel on every backend
2026-10-04 · commit e06ddc0 · feat(review)!: remove RepoPrompt support
2026-10-02 · commit af27800 · fix(review): route the unattended NEEDS_HUMAN escalation through the fix pass (B8)
2026-10-02 · commit d1b1db0 · fix(review): an unattended reviewer NEEDS_HUMAN on a non-blocking call becomes an open item (B8)
2026-10-01 · commit 282f63d · revert(working-rules): restore the measured attended background-review wording
2026-10-01 · commit 1fe6d1f · fix: group C wording gaps and audit autofix PR
2026-10-01 · commit 6b188cc · fix: review findings on the progressive-disclosure moves
2026-10-01 · commit 5df41db · perf(impl-review): codex fix pass loads only on NEEDS_WORK
2026-09-30 · commit 936493a · fix(impl-review): task ids, backend-owned verdicts, env opt-ins
2026-09-30 · commit 55a5192 · fix(skills): final-pass consistency for the review loop and PR wording
2026-09-30 · commit 99d1f1c · feat(review): unattended runs loop review until SHIP; attended keeps one re-review
2026-09-30 · commit 44b6c82 · fix(impl-review): finalize the fan-out before changing anything
2026-09-29 · commit 0b93768 · fix(impl-review): the review block defaults to three reviewers; one reviewer is the explicit opt-in
2026-09-29 · commit d45ffdc · feat(7.0): review panel by risk: persisted or shared state gets three reviewers whatever the size
2026-09-29 · commit 7e3da37 · feat(7.0): review loops until SHIP only on --until=merge or when asked
2026-09-29 · commit 9b2399f · feat(7.0): re-reviews see the author's declined findings and may withdraw them
2026-09-29 · commit f98f1fb · fix(impl-review): every hot-path block carries the full FLOWCTL ladder (codex mirror check)
2026-09-29 · commit 47ef951 · feat(7.0)!: impl-review hot path: the attended codex review lives in SKILL.md
2026-09-29 · commit 6ebddb0 · feat(7.0)!: impl-review fixes once and re-reviews once
2026-09-29 · commit 55729da · refactor(7.0): every stage skill reads working-rules.md; drop restated rules
2026-09-29 · commit 5ee77b2 · feat!: remove Ralph
2026-09-26 · commit b4c980a · Unattended-run correctness fixes (audit wave 1) (#470)
2026-09-11 · commit f92da65 · flow: attended conductor, shared routing reference, opinionated defaults (#416)
2026-09-05 · commit cfc29f1 · claude CLI review backend: claude -p as a packaged cross-family reviewer (#403)
2026-09-04 · commit 6f7990d · docs: model ids as placeholders in examples; dated reach notes for Fable 5.1, GPT-6 Astra and the Cursor wind-down (#400)
2026-09-02 · commit 7bf525d · Parallel-draw review: three axis lenses, merge, fix once (#392)
2026-08-27 · commit cd9664b · Issue sweep: excused review status, land tail, export roster, closers (#372)
2026-08-09 · commit a816353 · style(review-family): failure-signature declaratives, Done-when bounds, intra-skill dedupe
2026-08-08 · commit d9804f4 · refactor(impl-review): branch-disclosure — leaner reached paths within the backend split
2026-08-03 · commit ddcda16 · Convergence-aware review terminals and per-surface blocking calibration (#290)
2026-08-02 · commit e34abed · feat(codex): honor CODEX_HOME for alternate homes; raise review cap 4 -> 8 (#289)
2026-07-30 · commit ac7c93f · Structured review-artifact schema in receipts (#273)
2026-07-28 · commit 01c33ac · Opus 5 autonomous seam retune (#247)
2026-07-26 · commit ae134b2 · fix(flow-next): stop the sandbox error teaching agents to widen the reviewer sandbox (#242)
2026-07-24 · commit 6738bc4 · docs(flow-next): route default planner and quality tiers to Claude Opus 5 (#238)
2026-07-24 · commit e764b0b · Reached-path skill prompt optimization (#237)
2026-07-22 · commit a5a4ea2 · Grok Build host fidelity: GROK_AGENT detection + host profile (fn-126) (#233)
2026-07-22 · commit 8ba018c · Cursor first-class experience: team-marketplace install + host-aware orchestration (fn-123) (#231)
2026-07-22 · commit 7d99b59 · flowctl hardening and performance completion sweep (#229)
2026-07-21 · commit e77395d · fix: post-series review sweep - 4 valid findings from merged PRs #221-#224 (#226)
2026-07-20 · commit f01881a · flowctl round-trip diet: batched reads and writes for hot skills (#218)
2026-07-09 · commit 4b30937 · fix(flow-next): review-loop runaway root fix — honest verdicts, convergence ratchet, deterministic cap (fn-90) (#202)
2026-07-04 · commit dd0ba47 · docs(flow-next): fix impl-review SKILL.md reference to moved opt-in phases (fn-87 R9 follow-up)
2026-07-04 · commit 4fb3089 · perf(flow-next): impl-review opt-in phases → on-demand file (fn-87 R9)
2026-07-04 · commit 031a005 · fix(flow-next): impl-review fix-loop holes — MAJOR_RETHINK + cap reset (fn-87 R3)
2026-07-02 · commit a81770d · refactor(skills): skill prompt diet — progressive-disclosure gating, dedupe, archaeology (fn-82, −10.7k always-loaded tok) (#193)
2026-07-02 · commit 08ade80 · docs(skills): skill runtime token plumbing — single-emission writes, round-trip elimination, fix-loop guards (fn-81) (#191)
2026-07-02 · commit a333703 · docs(skills): RP_ELIGIBLE gate for impl-review & spec-completion-review steering (fn-80, fn-78 fast-follow) (#190)
2026-07-01 · commit a7297f9 · feat(review): cursor backend + no-embed + sharper prompts + per-task routing — flow-next 2.5.0 (#184)
2026-06-05 · commit d39f75e · feat(platform,tracker): Cursor local-install (mac/linux/windows) + tracker create-if-unlinked (#162)
2026-05-26 · commit 54269f7 · fn-48: Backend-split review workflows + FLOWCTL prelude consolidation (#146)
2026-04-24 · commit a814e5c · feat(flow-next): opt-in review flags --validate/--deep/--interactive — 0.35.0 (#118)
2026-04-24 · commit a5e2aef · feat(flow-next): review rigor bundle (R-IDs, confidence, classification, protected artifacts, triage-skip) — 0.32.1 (#115)
2026-04-24 · commit 295c3d4 · feat(flow-next): codex default gpt-5.5 + Codex browser rename (v0.32.0)
2026-04-22 · commit 59eeb68 · feat(flow-next): cmd_review_backend + skills + Ralph accept spec form (fn-28.4)
2026-04-22 · commit f519faa · chore(flow-next): default copilot model gpt-5.2 + high
2026-04-22 · commit 18840a8 · feat(flow-next): review skills branch on copilot backend
2026-02-03 · commit 72068df · fix(flow-next): cross-platform variable support for Claude Code + Droid
2026-02-03 · commit 292f865 · fix(flow-next): strengthen workflow.md directive in review SKILL.md files
2026-02-03 · commit a816864 · fix(flow-next): remove duplicate RP workflow code from SKILL.md files
2026-01-31 · commit bd7b6eb · fix(flow-next): hide duplicate skills from menu via user-invocable: false
2026-01-24 · commit f0815ec · fix(flow-next): restore manual prompt building for RP reviews (#73)
2026-01-23 · commit be6b960 · fix(flow-next): add WARNING block to prevent skipping verdict follow-up
2026-01-23 · commit 077f45a · fix(flow-next): clarify verdict extraction - ignore RP format, use follow-up only
2026-01-23 · commit a246a23 · chore(flow-next): add stronger workflow.md references in SKILL.md
2026-01-23 · commit efe0329 · fix(flow-next): add verdict follow-up step to impl-review and plan-review SKILL.md
2026-01-23 · commit c8240b9 · fix(flow-next): add --response-type review to SKILL.md setup-review example
2026-01-21 · commit d0a4e6d · feat(flow-next): use RP 1.6.0 builder review mode for impl-review (#61)
2026-01-18 · commit d9a1a5d · fix(flow-next): auto-fix review issues without user prompt (#55)
2026-01-16 · commit 56159c7 · feat(flow-next): centralize review backend detection (#49)
2026-01-16 · commit 8358a59 · fix(flow-next): scope impl-review to current task's commits only
2026-01-11 · commit 3ad9053 · fix(flow-next): ralph-init codex detection + --mode→--review
2026-01-11 · commit 840e7d3 · fix(flow-next): remove hardcoded opus model from review skills (#10)
2026-01-11 · commit 6c6ec85 · fix(flow-next): add backend prompt to review skills for interactive mode
2026-01-11 · commit fd0db5c · feat(impl-review): add codex backend support
2026-01-10 · commit ede923d · release: flow-next 0.3.17, flow 0.8.4
2026-01-09 · commit 1f1d56e · fix(flow-next): strengthen flowctl bundled warning, bump 0.3.8
2026-01-09 · commit 375b637 · fix: $ARGUMENTS in all skills, bundled note in export-context
2026-01-09 · commit 47267c4 · fix: add bundled flowctl note to all skills
2026-01-09 · commit d5db6f1 · feat(flow-next): Ralph autonomous mode improvements
2026-01-09 · commit 3904c73 · refactor(flow-next): consolidate RALPH_MODE to FLOW_RALPH
2026-01-09 · commit 023a943 · refactor(flow-next): remove hooks, clean rp-cli refs from review skills
2026-01-09 · commit e8f95c6 · chore: init
2026-01-09 · commit da6f148 · feat(flow-next): epic branches + clean flow commits
2026-01-09 · commit 7af7abc · feat(flow-next): atomic setup-review to prevent drift
2026-01-07 · commit 45f4bae · feat(flow-next): autonomous mode flags v0.2.0
2026-01-06 · commit 1770d0f · flow-next: prep-chat, review fixes, release READMEs
2026-01-06 · commit a9afcca · flow-next: new plugin with multi-user safety

Archived snapshots