|
|
@@ -26,7 +26,7 @@ From Python async patterns to Rust ownership models, from AWS Fargate deployment
|
|
|
- 📊 **`r-ops` skill** — the set's first data-science skill: a tidyverse-first, current-best-practice reference for modern R (2024+). `SKILL.md` routes an import → tidy → transform → visualize → model → communicate workflow across **9 reference files (~115 KB)** — tidyverse-core, import-io, strings-dates-factors, visualization, iteration-functional, modeling-stats, data-table, time-series, workflow-tooling. Leads with current idioms (native `|>`, dplyr `.by=`, the `\(x)` lambda, `across()`, `list_rbind`, `slice_*`, tidymodels, the tidyverts `tsibble`/`fable` stack, Quarto + renv) and names base R / `data.table` where they win. Ships a 43-assertion offline self-test plus a `check-r-facts.py` §7 staleness verifier (`--offline` asserts every catalogued CRAN package is still named in the prose and the currency note carries a year; `--live` resolves each package on CRAN) so the modern-stack claim is **machine-enforced, not asserted**. Salvaged and freshened from the stale stacked PR #6 (which also duplicated the already-shipped supply-chain-defense), re-landed clean off current `main`.
|
|
|
|
|
|
**v3.3.0** (June 2026)
|
|
|
-- 🔁 **`loop-ops` skill** — the *outer-loop* design discipline, twin to [`iterate`](skills/iterate/) (the inner loop). Where `iterate` drives one metric in one session, `loop-ops` is the orchestration layer above it: how to design, scaffold, cost, and **safely** run scheduled discover→triage→implement→verify→escalate-or-land agent loops. Its spine is the **risk-tier ladder** — L1 report → L2 assisted → L3 unattended — mapped onto Claude Code's *actual* permission model (the part a generic-agent methodology can't reach): each tier is a concrete permission mode, with the *enumerate-vs-isolate* fork and the load-bearing rule that **a scheduler invokes `claude -p`, not a session that spawns ungated children**. Ships a STATE/run-log/budget state spine, a **13-pattern catalog framed as a morphology** (trigger × posture × locus — incl. event-driven Channels and `/goal`-completion archetypes like metric-chase & backfill), multi-loop coordination + kill switch, and three Resource-Protocol scripts — `loop-scaffold` (scaffold), `loop-check` (readiness scorer that refuses a green light on an unbounded scope / missing gate / undefined escalation), `loop-estimate` (token-$ estimate by pattern × cadence × model). Composes `fleet-worker` (spawn) and `fleet-ops` (land); 109-assertion offline suite. Builds on the public *loop engineering* discipline (Steinberger, Osmani) and the [Ralph loop](https://ghuntley.com/ralph/), grounded in this repo's auto-mode-classifier reference.
|
|
|
+- 🔁 **`loop-ops` skill** — the *outer-loop* design discipline: how to design and **safely** run scheduled, autonomous agent loops — the orchestration layer above [`iterate`](skills/iterate/) (which drives a single inner run). Its backbone is a **graduated-autonomy risk ladder** (L1 report → L2 assisted → L3 unattended) mapped onto Claude Code's *actual* permission model, so a loop only ever gets the authority it has earned — anchored by the rule that **a scheduler invokes `claude -p`, never a session spawning ungated children**. Ships a **13-pattern morphology** — `trigger` (cadence · event-driven Channels · `/goal` run-to-completion) × `posture` (L1–L3) × `locus` (connector→cloud routine · local→Desktop task) — plus a STATE/run-log/budget spine, multi-loop coordination, and a kill switch. Three tools do the work: **`loop-scaffold`** seeds a near-ready loop, **`loop-check`** refuses a green light on an unbounded scope / missing gate / undefined escalation, and **`loop-estimate`** gives caching-aware cost/month before you commit to a cadence. Composes `fleet-worker` (spawn) and `fleet-ops` (land).
|
|
|
|
|
|
**v3.2.0** (June 2026)
|
|
|
- 🤖 **`fleet-worker` skill** — delegate tool-using, multi-step tasks to *cheaper headless Claude Code workers* — a cheaper Anthropic model (Sonnet/Haiku) or any Anthropic-compatible endpoint (e.g. GLM 5.2 via z.ai) — while an Opus orchestrator fans them out in parallel and gates their results before anything lands. Each worker is a real `claude -p` with Claude Code's full tool harness (Read/Write/Edit/Bash/Glob/Grep/Task) and any skills you provision into it, but a cheaper brain — isolated in its own git worktree + `CLAUDE_CONFIG_DIR`. Ships bash + PowerShell launchers, a result-gating collector, an endpoint health verifier, and the fleet-ops handoff recipes. fleet-worker is the **spawn** layer; [`fleet-ops`](skills/fleet-ops/) is the test-gated **landing** layer it hands winning branches to. Provider-agnostic.
|
|
|
@@ -37,7 +37,7 @@ From Python async patterns to Rust ownership models, from AWS Fargate deployment
|
|
|
- 📋 **`adr-ops` skill** - Architecture Decision Records as a cross-project workflow. ADRs are append-only project memory: they capture *why* a system took its shape — the alternatives weighed, the constraints accepted — so a future maintainer recovers the reasoning without archaeology through git history or chat logs. Brings the when-to-write rule, canonical format, proposed→accepted→superseded lifecycle, and append-only supersession discipline, with five Resource-Protocol tools (init / new / index / `touches`-query / lint) and a 72-assertion suite.
|
|
|
- 📚 **`okf-ops` skill** - assess, validate, and adopt the [Open Knowledge Format](https://cloud.google.com/blog/products/data-analytics/how-the-open-knowledge-format-can-improve-data-sharing/) — Google Cloud's vendor-neutral spec (v0.1, Apache-2.0) for packaging organizational knowledge as a directory of markdown files with YAML frontmatter that AI agents can query without a platform or SDK. A read-only readiness scanner finds good adoption candidates across many repos; a conformance validator (`--strict` for CI) checks a bundle. Honest scope baked in — OKF is a v0.1 draft, adopt per-repo.
|
|
|
- 📦 **`pypi-ops` skill** - publish Python packages to PyPI the 2026 way: OIDC Trusted Publishing with PEP 740 attestations via `gh-action-pypi-publish`, not stored API tokens. First-publish pending-publisher setup, the `invalid-publisher` / "already exists" failure ladder, TestPyPI dry runs, release-environment approval gates, local `uv publish` / `twine`, and a stale-OIDC-federation audit.
|
|
|
-- 🔍 **github-ops auditor family** - a read-only repo-health suite: a security-posture auditor (Dependabot / secret + code scanning / private vuln reporting / SECURITY.md / branch protection, visibility-aware severity), open-issue surfacing wired into the pre-push gate as an advisory, and a scored `repo-scorecard` capstone that grades a repo — or an entire `--org` — in one pass, emitting fix commands but never applying them. The whole family now renders through the `term.sh` panel design system.
|
|
|
+- 🔍 **github-ops repo-health auditing** — know a repo's security posture at a glance, without leaving the terminal. A **read-only** auditor checks the controls that actually matter (Dependabot, secret + code scanning, private vulnerability reporting, `SECURITY.md`, branch protection) with visibility-aware severity, flags stale open issues as a pre-push advisory, and rolls it all into a scored **`repo-scorecard`** that grades a single repo — or an entire `--org` — in one pass. It emits the exact fix commands and never touches your repo, so you decide what to apply.
|
|
|
|
|
|
**v3.0.0** (June 2026)
|
|
|
- **Skills-first restructure** - *Breaking:* the expert-agent layer was cut from 23 to 3. Per Anthropic's guidance, knowledge belongs in skills (progressive disclosure, single source of truth) and subagents are reserved for context isolation — so *all* domain-knowledge agents became `-ops` skills (the 11 language/framework experts → their twins; cypress/cloudflare/bash/craftcms/payloadcms/asus-router → new skills; claude-architect/aws-fargate folded into existing skills). The 3 remaining agents are pure isolation/worker roles: `git-agent` (background commits/PRs), `firecrawl-expert` (noisy multi-page scrapes), `project-organizer` (bulk restructure). Dispatching skills now route `general-purpose` agents that preload skill references.
|
|
|
@@ -285,7 +285,7 @@ See [skill-creator](skills/skill-creator/) for the complete guide.
|
|
|
|-------|-------------|
|
|
|
| [tool-discovery](skills/tool-discovery/) | Recommend agents and skills for any task |
|
|
|
| [git-ops](skills/git-ops/) | Git orchestrator - commits, PRs, releases, changelog. Routes to background Sonnet agent. |
|
|
|
-| [github-ops](skills/github-ops/) | GitHub remote operations - repo creation, releases, metadata, README Recent Updates convention |
|
|
|
+| [github-ops](skills/github-ops/) | GitHub remote ops - repo create/metadata/topics, releases + README 'Recent Updates' enforcement, issue/PR management (preview-before-send), and read-only security-posture audit + scored repo-scorecard (single repo or whole `--org`) |
|
|
|
| [push-gate](skills/push-gate/) | Pre-push safety gate - gitleaks + regex secret scan, forbidden-file check, no bypass |
|
|
|
| [fleet-ops](skills/fleet-ops/) | Manage a fleet of concurrent Claude sessions - landing queue with test gate, pre-land scrub (experimental) |
|
|
|
| [fleet-worker](skills/fleet-worker/) | Delegate tasks to cheap headless GLM (or any Anthropic-compatible) workers - per-task git worktree + isolated config, result gating, fan-out that hands winning branches to fleet-ops landing |
|
0xDarkMatter