review-new-claude-features
Skill9 starsChanged 7 months agoMITView on GitHub
Scout the latest Claude Code / Anthropic platform releases and decide, against THIS hub's adoption doctrine, which new features to ADOPT, KEEP-hand-rolled, REJECT, or MEASURE-FIRST — then file migration issues for the adopt/measure candidates. The FREE, on-demand, in-session form of the platform-migration self-updating loop (Layer 1): run it in a session you are already in — no scheduled cloud agent, no recurring spend. Use when you want to check "what new Claude Code features should this hub adopt?" after a CC release, periodically, or before planning hub work. NOT for general external pattern scanning (use /self-improve), and NOT the paid scheduled-routine version (tracked separately as a recurring-spend owner decision — see issue #153).
In this repository44 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- anthropic-multi-agent-research-system-skillSkill
- apply-selectionsSkill
- bootstrap-dogfood-projectSkill
- brainstormSkill
- branch-choiceSkill
- cc-adoption-scoutSkill
- claude-guardianSkill
- continueSkill
- end-sessionSkill
- executing-plansSkill
- five-advisorsSkill
- get-work-doneSkill
- git-branch-lifecycleSkill
- githubSkill
- grill-meSkill
- hook-transcript-fixture-testSkill
- instagramSkill
- linkedinSkill
- modern-data-pipeline-diagramSkill
- pattern-qualitySkill
- platform-event-live-probeSkill
- plugin-lifecycleSkill
- promote-to-coreSkill
- prompt-auto-enhanceSkill
- provision-reportSkill
- redditSkill
- save-sessionSkill
- scan-discovery-reportSkill
- scan-repoSkill
- scan-urlSkill
- self-improveSkill
- skill-evaluatorSkill
- skill-masterSkill
- ssot-workflow-auditSkill
- start-sessionSkill
- synthesize-hubSkill
- synthesize-projectSkill
- twitter-xSkill
- verify-effect-at-destinationSkill
- workflow-doc-reviewerSkill
- writing-plansSkill
- writing-skillsSkill
- youtube-transcriptSkill
The file
Show its rules and text
We split this skill into the rules it gives an agent
15 rules. About 0 are common, 7 are distinctive, and 8 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈7
Distinctive · estimate
This author's own choices. Worth a look.
8
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 46
Read the hub's adoption doctrine so judgments are anchored, not ad-hoc:
- Very specific: line 51
Read the wired release-tracking URL entries from config/urls.yml (a flat list of url: entries, added in Phase 5.1a) — the official Claude …
- Distinctive: line 88
Before filing anything, check it is genuinely new:
- Very specific: line 89
Search the plan ledger + registry/changelog.md for an existing adopt/keep/reject record.
- Very specific: line 90
Search open + recently-closed GitHub issues (gh issue list --search) so you comment on an existing tracking issue instead of duplicating …
- Very specific: line 93
If scan_web/discovery_to_issue already filed a discovery for it, ENRICH that issue with the verdict rather than opening a new one.
- Distinctive: line 98
Always print a verdict table: feature · version · status · verdict · one-line reason.
- Very specific: line 99
--dry-run(default): stop here — report only, file nothing. - Very specific: line 110
MUST run in-session only — MUST NOT create a scheduled cloud routine, invoke any billed/ ultra feature, or spend money.
- Distinctive: line 112
MUST anchor every verdict to the hub doctrine (reactive-not-speculative · KISS/YAGNI · measure-before-adopt · additive-for-downstream), …
- Very specific: line 114
MUST quote the official doc verbatim for any GA/availability/version claim — never paraphrase it (the conflation that caused a wrong …
- Distinctive: line 116
MUST classify a spend- or policy- or default-changing feature as PROPOSE-only and file it for owner decision — never auto-enact.
- Distinctive: line 118
MUST dedupe against the plan ledger + existing issues before filing.
- Very specific: line 120
MUST default to --dry-run.
- Distinctive: line 121
MUST advance the reviewed-baseline marker so successive runs don't re-audit settled releases.
≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
/review-new-claude-features — free in-session Claude Code feature-adoption judgment
This is the agentic judgment half of the hub's self-updating Layer 1. The mechanical
half already exists and is free: scan-internet.yml (weekly) → scan_web.py →
discovery_to_issue.py surface new-feature discoveries as issues. What was still manual
is the JUDGMENT — "should this hub adopt / keep / reject / measure this feature?" — which
is exactly the work the 2026-06-20 platform-migration session did by hand. This skill makes
that judgment a repeatable, free, on-demand procedure (runs in your current session;
costs only the usage of the session you are already in — no scheduled cloud agent).
Cost contract (load-bearing): this skill MUST NOT create a scheduled cloud routine, call any billed/
ultrafeature, or spend money. It runs in-session only. The paid scheduled-routine version is a separate owner decision (issue #153) — never auto-enact it.Owner-approved standing invocation (2026-08-09): the GetWorkDone keeper runs this skill WEEKLY, headless, on the VPS (
GWD\feature-adoption-sweep.ps1, self-gated marker + IST delivery window; verdict card delivered vianotify-owner.ps1; baseline advanced inheartbeats\feature-sweep-baseline.txt, default 2.1.226). This is plan-usageclaude -pon owned hardware — NOT the billed cloud routine #153 gated on — and satisfies the contract: the OWNER scheduled it; the skill still never schedules itself.
Arguments: $ARGUMENTS — default is --dry-run (report only; file nothing).
STEP 1: Load the doctrine + the release sources
- Read the hub's adoption doctrine so judgments are anchored, not ad-hoc:
plans/platform-migration-2026H2.md— the ledger + the adopt/keep/reject/measure precedents.core/.claude/rules/rule-curation.md(reactive-not-speculative),claude-behavior.mdrules 16/21/22 (KISS / YAGNI / measure-before-optimize),agent-orchestration.md(single-level + the S2 "adopt only if measured better" precedent).
- Read the wired release-tracking URL entries from
config/urls.yml(a flat list ofurl:entries, added in Phase 5.1a) — the official Claude Code docs:whats-new,sub-agents,skills,hooks,worktrees,scheduled-tasks,agents-and-tools(+ any added since). - Resolve the baseline: the last-reviewed CC version. Default = the most recent version
named in
plans/platform-migration-2026H2.md; override with--since vX.Y.Z.
STEP 2: Fetch what shipped since the baseline (adversarial, not single-source)
Fetch the current release docs with WebFetch (or dispatch parallel web-research workers via
Agent() at T0 for breadth). For higher-stakes verification, prefer the native /deep-research
harness so a claim is cross-checked, not taken from one page. Extract every feature/primitive
new since the baseline — name, ship version, GA-vs-experimental status, and the exact
doc quote (never paraphrase a GA/availability claim — the 2026-06-20 audit caught a
"research-preview vs GA" conflation that cost a wrong deferral).
STEP 3: Judge each feature against the doctrine (the core step)
For EACH new feature, classify into exactly one bucket, with a one-line reason:
| Verdict | When | Example precedent |
| ADOPT | A concrete hub caller benefits AND it beats the hand-rolled version (KISS win, real capability) | SubagentStart/ConfigChange hooks (verified live) |
| MEASURE-FIRST | Plausible benefit but unproven for a hub workflow — needs an A/B before adopting | Dynamic Workflows tool → S2 A/B |
| KEEP hand-rolled | The native primitive exists but the hub's version is equal/better or migrating is churn | Routines vs the deterministic GH-Action crons (negative ROI) |
| REJECT / not-ready | Experimental, flag-gated, or breaks a hub guarantee | Agent Teams (experimental, breaks session resumption) |
Hard rules for the judgment (these ARE the doctrine):
- Reactive, not speculative — ADOPT only with a concrete caller; never "because it's new."
- Measure before adopting a feature whose benefit is a performance/parallelism claim (the S2 precedent: identical wall-clock → not adopted).
- Additive / opt-in for anything downstream inherits — never change a distributable default without flagging it as an owner decision.
- Spend / policy = owner decision — a feature that costs money (billed review, scheduled cloud agent) or changes execution policy is PROPOSE-only; file it, never enact it.
STEP 4: Dedupe against what the hub already decided
Before filing anything, check it is genuinely new:
- Search the plan ledger +
registry/changelog.mdfor an existing adopt/keep/reject record. - Search open + recently-closed GitHub issues (
gh issue list --search) so you comment on an existing tracking issue instead of duplicating (reuse thediscovery_to_issue.py//create-github-issuededup conventions). - If
scan_web/discovery_to_issuealready filed a discovery for it, ENRICH that issue with the verdict rather than opening a new one.
STEP 5: Report — and file issues only with --apply
- Always print a verdict table: feature · version · status · verdict · one-line reason.
--dry-run(default): stop here — report only, file nothing.--apply: for each ADOPT or MEASURE-FIRST verdict, file (or comment on) a migration issue linked to the platform-migration epic, with: the doc quote, the verdict + reason, a migration sketch, and an explicitowner-decisionflag if it implies spend or a policy/default change. KEEP and REJECT verdicts are recorded in the report only (no issue) — unless a KEEP overturns a prior ADOPT, which warrants a tracking note.- Append a one-line entry to
plans/platform-migration-2026H2.md's log so the baseline advances and the next run does not re-audit the same release.
CRITICAL RULES
- MUST run in-session only — MUST NOT create a scheduled cloud routine, invoke any billed/
ultrafeature, or spend money. The paid scheduled version is owner-gated (issue #153). - MUST anchor every verdict to the hub doctrine (reactive-not-speculative · KISS/YAGNI · measure-before-adopt · additive-for-downstream), never "adopt because new".
- MUST quote the official doc verbatim for any GA/availability/version claim — never paraphrase it (the conflation that caused a wrong deferral on 2026-06-20).
- MUST classify a spend- or policy- or default-changing feature as PROPOSE-only and file it for owner decision — never auto-enact.
- MUST dedupe against the plan ledger + existing issues before filing; enrich an existing discovery issue rather than duplicating it.
- MUST default to
--dry-run; file issues only when--applyis explicitly passed. - MUST advance the reviewed-baseline marker so successive runs don't re-audit settled releases.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

