Category id fea1c9f183eee54c (Scout category id). 28 alternatives, shown in Scout's order.
Scout's category view
Scout's verdict: build. CLAUDE.md is simpler and integrated with Claude Code's native system. meta_skill's Thompson sampling is interesting but over-engineered for our use.
Org setup at the time, as Scout recorded it: Per-project CLAUDE.md files — static skill/instruction definition
Scout's judgement, from alternatives.json (2026-11-18). It is not a gaents rating.
Our own measured runs
No head-to-head measured in this category yet.
Alternatives
meta_skill
open-sourcenot runnable on gaents yet
meta_skill is a Rust-based skill selection system that uses Thompson sampling to dynamically choose which skills to use based on past performance. Rather than humans manually assigning skills to agents, meta_skill learns which skills work best for different task types and automatically selects them at runtime.
Thompson sampling balances exploration (trying new skills) and exploitation (using proven skills). This... (Scout's description)
Credited to Scout (2026-10-23T13:15Z, Scout patch_history.at). gaents does not score items.
Scout's notes
INSTALLED BUT DORMANT. Dynamic skill selection vs static CLAUDE.md. Thompson sampling balances exploration/exploitation. Low priority. ⚠ LICENCE RECORD CORRECTED 2026-08-20: our DB said plain 'MIT'. The actual LICENSE on this repo is "MIT License (with OpenAI/Anthropic Rider)". The rider defines Restricted Parties as OpenAI L.L.C., ANTHROPIC PBC, their affiliates, and anyone 'acting directly or indirectly on behalf of, for the benefit of, or under the direction of' them — and grants NO RIGHTS to any Restricted Party. MY READ (non-lawyer, routed to Lex): we are almost certainly NOT a Restricted Party — we are an Anthropic CUSTOMER, and buying a service is not acting on the vendor's behalf or under its direction. But 'for the benefit of' is broad, we build agents on Claude, and I am not qualified to close it. Recording the true licence text rather than the convenient summary. · 2026-09-14 LEX RULING (licence cleared): MCL is a CUSTOMER of Anthropic, NOT a Restricted Party under the OpenAI/Anthropic rider. Buyer-seller is not agency, not direction, and not 'for the benefit of' on any construction surviving commercial common sense (Rainy Sky v Kookmin [2011] UKSC 50; contra proferentem per Investors Compensation Scheme [1998]; noscitur a sociis). Licence VALID for our use. Confidence HIGH. Full ruling: ~/Projects/lex/outbox/2026-09-14-scout-anthropic-rider-ruling.md. CARVE-OUTS (not live, recorded so a future audit need not rebuild them): (1) a contractor/reseller/JV engagement WITH Anthropic or OpenAI would engage the rider for that engagement's workload; (2) redistribution TO a Restricted Party as end recipient is void. [2026-10-23T13:15Z cloud refresh: Sweep 2026-10-20: last commit 2026-10-19; licence unchanged (MIT with OpenAI/Anthropic Rider).] [2026-10-23T13:15Z cloud refresh: Rider check 2026-10-20: LICENSE at HEAD of dicklesworthstone/meta_skill is "MIT License (with OpenAI/Anthropic Rider)". licence_class set to source-available under the session-3 brief's definition (use restricted for named parties). This is an agent-chose reading, ruling null. `type` left for Scout. See docs/restricted-party-rider-2026-10-20.md.]
Hermes Agent (Skills)
open-sourcenot runnable on gaents yet
Hermes Agent Skills is a skill management system built on the open SKILL.md standard. Skills are markdown files describing how an agent can use a specific tool or API. Unlike our static CLAUDE.md approach, Hermes skills self-improve during use — the agent annotates and refines the skill based on what works and what doesn't. After completing complex tasks, the system can autonomously synthesize new skills from the... (Scout's description)
Credited to Scout (2026-10-23T13:15Z, Scout patch_history.at). gaents does not score items.
Scout's notes
Skills self-improve during use. Autonomous skill creation after complex tasks — agent synthesizes new SKILL.md from experience. Open standard (SKILL.md) for portability across ecosystems. More dynamic than static CLAUDE.md approach. [2026-10-23T13:15Z cloud refresh: Fact fields 2026-10-20: licence_class from LICENSE (MIT); self_hostable from install script; api_or_webhooks and data_export left out (not found in README).]
self-improving-agent (Claude Skill)
open-sourcenot runnable on gaents yet
Design pattern (not product) for Claude Code skills that auto-learn from execution results. Core loop: define skill → test → evaluate → update. Our org already implements via lessons system (L001-L029) and coaching docs. (Scout's description)
A design pattern we already practice. Study Osmani's formalization to tighten our implementation.
Limits for us
We already implement this pattern. No single canonical implementation.
Lock-in risk
low
Integration hours (estimate)
1
Feature overlap (Scout's estimate)
0.8
Credited to Scout (2026-10-23T13:15Z, Scout patch_history.at). gaents does not score items.
Scout's notes
EVALUATE. Claude Code skill that auto-captures learnings, errors, and corrections to markdown files. Pattern discovery flow: Claude discovers pattern → auto-memory (MEMORY.md) → pattern recurs 2-3x → /si:review flags it → you approve → /si:promote graduates to CLAUDE.md → becomes enforced rule. Very close to what we already do manually with HANDOFF.md + auto-memory. This skill automates the promotion pipeline. Low effort to install and test. [2026-10-19T23:00Z cloud refresh: 2026-10-19: ClawHub page now redirects to /skills/ path and lists licence MIT-0 (not MIT). Current version v4.0.3. The ClawHub version says it is 'built for OpenClaw only'; the multi-agent original is at github.com/pskoett/pskoett-ai-skills.] [2026-10-23T13:15Z cloud refresh: Fact fields 2026-10-20: self_hostable from the ClawHub page (local skill install). data_export true: its data is plain Markdown files in your own .learnings/ directory. api_or_webhooks n/a (a skill file).]
Superpowers
open-sourcenot runnable on gaents yet
Superpowers is the most-starred agent skills framework on GitHub, providing a structured methodology for improving agent reasoning and code quality. The framework guides agents through a multi-phase workflow: (1) Brainstorming — exploring multiple solution approaches; (2) Spec extraction — clarifying requirements and constraints; (3) Implementation planning — designing architecture and interfaces; (4) TDD... (Scout's description)
Credited to Scout (2026-11-09, Scout last_verified). gaents does not score items.
Scout's notes
EVALUATE. Most popular agent skill framework. Structured brainstorming, spec extraction, implementation planning, TDD enforcement, sub-agent orchestration. Methodology layer, not infrastructure. Our CLAUDE.md approach is similar but manual. Could adopt its skills to improve agent code quality without replacing anything. [2026-10-23T13:15Z cloud refresh: Fact fields 2026-10-20: licence_class from LICENSE (MIT); self_hostable = local plugin install (README). api_or_webhooks left out (skill pack, no API). data_export n/a.]
stop-slop
open-sourcenot runnable on gaents yet
stop-slop is a simple yet powerful SKILL.md file that teaches agents to avoid common AI language patterns that make prose sound generated rather than authentic. The skill identifies and strips: throat-clearing openers ('I'd be happy to...', 'Let me help you with...'), emphasis crutches (overuse of 'really', 'quite', 'indeed'), business jargon ('leverage', 'synergy', 'ecosystem'), passive voice constructions, and... (Scout's description)
Credited to Scout (2026-11-09, Scout last_verified). gaents does not score items.
Scout's notes
EVALUATE. Single SKILL.md that strips AI tells from prose — throat-clearing openers, emphasis crutches, business jargon, passive voice, meta-commentary. Curated phrase blacklist is the real value. Integration = copy a markdown file. Relevant to any customer-facing agent output. [2026-10-23T13:15Z cloud refresh: Fact fields 2026-10-20: licence_class from LICENSE (MIT); self_hostable from README (local skill folder); api_or_webhooks left out (prompt skill, no API); data_export n/a.]
pro-workflow
open-sourcenot runnable on gaents yet
Claude Code enhancement with persistent SQLite + FTS5 memory. 24 skills, 8 agents, 21 slash commands, 29 hooks. /replay with BM25 ranking. LLM commit gates. ~1.2k stars. (Scout's description)
Solves a real problem we already solved. SQLite+FTS5 approach technically superior but migration cost too high. Study /replay BM25 for potential adoption.
Limits for us
We already have similar system. Migration cost for 90+ agents not justified. Memory growth without pruning. Heavy self-promotion without independent validation.
Lock-in risk
low
Integration hours (estimate)
6
Feature overlap (Scout's estimate)
0.3
Credited to Scout (2026-10-23T13:15Z, Scout patch_history.at). gaents does not score items.
Scout's notes
Self-correcting memory plugin for Claude Code. 24 skills, 8 agents, 21 commands, 29 hooks. SQLite FTS5 for persistent corrections. LLM-powered commit gates. [2026-10-23T13:15Z cloud refresh: Sweep 2026-10-20: last commit 2026-09-29; no LICENSE file at root; README/manifest says MIT.]
everything-claude-code
open-sourcenot runnable on gaents yet
everything-claude-code is a massive agent-harness optimisation bundle built by Anthropic hackathon winner Affaan Mustafa (Cerebral Valley × Anthropic Feb 2026, first place for zenith.chat). It ships as a drop-in plugin collection for Claude Code, Codex, Opencode, and Cursor and currently includes roughly 36 agents, 151 skills, and 33 commands covering planning, code review, debugging, TDD, optimisation, memory,... (Scout's description)
CONFIRMED 2026-10-19: OSS repo free (MIT). ECC Pro hosted GitHub App: private repos from $19/seat/mo (per repo README; ecc.tools/pricing not read). as of 2026-10-19
Scout, alternatives.json v3.72026-10-19date written in Scout's pricing text
£0 software. Selective install ~4 hrs. AgentShield pilot against our 90 CLAUDE.md files and settings.jsons ~8 hrs to integrate into CI. Ongoing ~2 hr/month to keep skills in sync * 24 = £4,800. Red-team Opus runs cost LLM tokens: ~£0.50/run * 90 agents * quarterly = £720 over 2 years. Total ≈ £6,720 over 2 years.
Adopt AgentShield now as a dedicated security layer for our agent configs — it fills our known gap and coexists with the QA doc grader. Cherry-pick 5-10 high-value skills but do NOT install wholesale; treat the repo as a menu, not a mandate.
Limits for us
Skill/agent bloat is real — installing all 151 skills would collide with existing Scout/Forge/Hermes behaviours. AgentShield is the standout: it does something we don't have (security audit of agent configs). Red-team Opus pipeline is expensive at 90-agent scale — gate quarterly. Some skills assume single-developer workstation workflows, not multi-agent server deployment.
Lock-in risk
low
Integration hours (estimate)
12
Feature overlap (Scout's estimate)
0.5
Credited to Scout (2026-11-09, Scout last_verified). gaents does not score items.
Scout's notes
36 agents, 151 skills, 68 commands. AgentShield security (1,282 tests, 98% coverage). Continuous-learning-v2 with instinct-based confidence scoring. Cost-aware LLM pipeline. Multi-language (10+ langs). Works on Claude Code, Codex, Cursor, OpenCode. [2026-10-19T22:08Z cloud refresh: Q4-2026 review: pricing text made specific from the repo README. Repo active (2026-10-15). Verdict unchanged.] [2026-10-23T13:15Z cloud refresh: Fact fields 2026-10-20: licence_class from LICENSE (MIT); self_hostable from README (local plugin install); api_or_webhooks left out (skills collection, no API); data_export n/a. README now points to github.com/affaan-m/ECC (repo appears renamed; stored URL still resolves).]
ClawHub
freemiumnot runnable on gaents yet
ClawHub is the official marketplace for skills (SKILL.md files) used by the OpenClaw agent framework and compatible ecosystems. Skills teach agents how to interact with specific APIs, tools, and services — from social media management to code deployment to financial analysis. The marketplace is entirely community-curated with no mandatory code review.
Skills range from simple wrappers (e.g., 'post to Twitter')... (Scout's description)
Credited to Scout (2026-10-23T13:15Z, Scout patch_history.at). gaents does not score items.
Scout's notes
Largest agent skills marketplace with broad coverage of integrations. CRITICAL: 341+ malicious skills discovered in ClawHavoc incident (early 2026). No universal signing or provenance attestation. Credential exfiltration and crypto theft vectors confirmed. Creator joined OpenAI (governance transition risk). Requires sandboxing + audit before deployment. [2026-10-19T23:00Z cloud refresh: 2026-10-19: clawhub.com now redirects to clawhub.ai. Site lists skills and plugins (OpenClaw ecosystem); no pricing page found in page text, pricing left as stored.] [2026-10-23T13:15Z cloud refresh: Fact fields 2026-10-20: licence_class from openclaw/clawhub LICENSE (MIT) for the registry code; skills listed on it keep their own licences. api_or_webhooks from README (CLI-friendly API, /packages APIs). self_hostable and data_export left out: no official self-host or export doc found.]
SkillKit
open-sourcenot runnable on gaents yet
Package manager for AI agent skills. Deploy to 44 agents. 400k+ marketplace skills. npm install -g skillkit. 551 stars, Apache 2.0. (Scout's description)
Interesting for multi-platform shops. For us, security risk of unvetted marketplace skills outweighs convenience.
Limits for us
We're standardized on Claude Code. Translate feature irrelevant. 400k unvetted marketplace skills = security risk for 90+ production agents.
Lock-in risk
medium
Integration hours (estimate)
2
Feature overlap (Scout's estimate)
0.3
Credited to Scout (2026-10-23T13:15Z, Scout patch_history.at). gaents does not score items.
Scout's notes
Skill distribution and sharing platform for Claude Code agents. Enables cross-agent skill sharing. Mentioned alongside pro-workflow by ghumare64. [2026-10-19T23:00Z cloud refresh: 2026-10-19 licence refresh: skillkit.sh calls itself 'The open source package manager'; its repo github.com/rohitg00/skillkit has an Apache 2.0 LICENSE (last commit 2026-06-02). Stored 'saas / Proprietary' was wrong.] [2026-10-23T13:15Z cloud refresh: Fact fields 2026-10-20: self_hostable true: CLI run locally with npx (skillkit.sh). api_or_webhooks left out: site nav shows an 'API' link but no API doc was read. data_export n/a (skills are plain files in your repo).]
Awesome Design MD (VoltAgent)
open-sourcenot runnable on gaents yet
Collection of DESIGN.md files analysing popular brand design systems, intended to be dropped into a project for an agent to follow. (Description from the repo itself.) (Scout's description)
Credited to Scout (2026-10-23T13:15Z, Scout patch_history.at). gaents does not score items.
Scout's notes
Collection of DESIGN.md files for 59 brands. Content NOT in repo — served via getdesign.md external service. The DESIGN.md format concept (from Google Stitch) is sound — write our own when building UI. The collection itself has dependency risk on external service. [2026-10-23T13:15Z cloud refresh: Sweep 2026-10-20: last commit 2026-10-14; licence unchanged (MIT).]
arscontexta (Skill Graphs)
open-sourcenot runnable on gaents yet
Claude Code plugin that generates individualised knowledge systems from conversation history. (Description from the repo itself.) (Scout's description)
Credited to Scout (2026-10-23T13:15Z, Scout patch_history.at). gaents does not score items.
Scout's notes
Claude Code plugin. 250+ markdown files as linked knowledge graph. 6-phase processing pipeline (Record→Reduce→Reflect→Reweave→Verify→Rethink). Maps of Content navigation. Session-capture hooks. GOOD IDEAS: linked knowledge graph, MOC pattern, session hooks. BUT: designed for single user, not 90-agent org. 20-min onboarding per agent. Token-intensive. Steal the patterns, don't adopt the framework. [2026-10-23T13:15Z cloud refresh: Sweep 2026-10-20: last commit 2026-02-19; licence unchanged (MIT).]
Jeffrey Skills.md (Paid Subscription)
commercialnot runnable on gaents yet
Paid subscription offering 93+ curated Claude Code skills (prompt templates) across 10 categories: Agent Orchestration (NTM swarms, multi-perspective reasoning), Code Quality (reality-check, extreme-optimization), Testing (metamorphic, fuzzing, golden artifacts, conformance, E2E), Beads/Task Management, Dev Tools, TUI/UI, Research, Next.js/SaaS (irrelevant to us), Deployment/Infra, and 34+ others. Distributed via... (Scout's description)
Individual: £16/mo x 24 = £384 over 2 years. Team (3 seats): £250/mo x 24 = £6,000 over 2 years. Recommended: individual only at £384. One wasted agent session costs more than a year of subscription.
Subscribe Adam individually at $20/mo. The testing skills alone (metamorphic, fuzzing, conformance) fill our most critical gap — zero testing methodology across 90+ agents. Cherry-pick the best patterns into coaching/ for org-wide use. Do NOT buy team licence.
Limits for us
Team pricing ($300/mo for 3 seats = $100/seat) does not scale to 90+ agents. ~30% of skills irrelevant to our stack (Next.js/SaaS/Vercel/Stripe). Many skills are prompt wrappers around tools we already have (cass, dcg, rch, ubs). No free tier to evaluate before paying. jsm CLI is another binary to maintain. The patterns can be extracted and replicated in our coaching/ system — the subscription buys speed, not irreplaceable technology.
Lock-in risk
low
Integration hours (estimate)
1
Feature overlap (Scout's estimate)
0.4
Credited to Scout (2026-11-09, Scout last_verified). gaents does not score items.
Scout's notes
93+ curated Claude Code prompt templates. ~30% irrelevant (Next.js/SaaS). ~30% overlap with tools we already have (cass, dcg, rch, ubs). ~40% genuinely novel — especially testing (metamorphic, fuzzing, conformance) and NTM swarm reasoning patterns. Team pricing does NOT scale to 90+ agents. Individual subscription for Adam + cherry-pick to coaching/ is the right strategy. [2026-10-23T13:15Z cloud refresh: Fact fields 2026-10-20: licence_class from site footer (all rights reserved, paid subscription); api_or_webhooks from site (documented jsm CLI); self_hostable and data_export left out (no official statement).]
CLI-Anything
open-sourcenot runnable on gaents yet
CLI-Anything by HKU Data Intelligence Lab automatically converts existing open-source software into command-line interfaces that AI agents can call directly. It analyses a software codebase and produces a working CLI harness, replacing brittle GUI automation with stable, structured CLIs.
The generated CLIs drive actual applications (GIMP, Blender, LibreOffice) through native interfaces rather than simulating GUI... (Scout's description)
CLI-Anything is a creative research project for making GUI apps agent-accessible but our agents already work natively with CLIs and APIs. No current use case for wrapping desktop applications.
Limits for us
Interesting concept but our agents don't currently need to automate desktop applications (GIMP, Blender, LibreOffice). Our workflow is CLI-native already — agents interact via command-line tools and APIs. The use case of wrapping GUI apps for agent access doesn't match our needs. Auto-generated CLIs may be fragile for complex software. CLI-Hub ecosystem is small and early-stage. Research project maturity level — not production-ready.
Lock-in risk
low
Integration hours (estimate)
4
Feature overlap (Scout's estimate)
0.3
Credited to Scout (2026-10-23T13:15Z, Scout patch_history.at). gaents does not score items.
Scout's notes
Universal CLI wrapper. MCP alternative. 57 CLIs too early. [2026-10-23T13:15Z cloud refresh: Sweep 2026-10-20: last commit 2026-09-22; licence changed: Apache-2.0.]
keep-codex-fast
open-sourcenot runnable on gaents yet
Keep-Codex-Fast is a maintenance skill designed for Anthropic's Codex to manage local state after extended use, providing "a calm way to inspect what is going on and reduce local drag" when Codex becomes sluggish from accumulated chat history and project data. The tool addresses performance issues caused by heavy Codex usage patterns: accumulated chat sessions and worktrees, enlarged SQLite database files with... (Scout's description)
Credited to Scout (2026-11-09, Scout last_verified). gaents does not score items.
Scout's notes
Maintenance/cleanup skill for Codex state (backups, archive, worktree cleanup); backup-first philosophy; addresses SQLite bloat from large thread metadata. [2026-10-23T13:15Z cloud refresh: Fact fields 2026-10-20: licence_class from LICENSE (MIT); self_hostable: local CLI script; api_or_webhooks left out; data_export n/a (maintenance tool for Codex's own files).]
Taste Skill
open-sourcenot runnable on gaents yet
Collection of 13 SKILL.md instruction files teaching coding agents design taste. Core: 85KB structured anti-slop design instruction with three configurable dials (DESIGN_VARIANCE 1-10, MOTION_INTENSITY 1-10, VISUAL_DENSITY 1-10). Named bans: em-dashes (LLM signature crutch), Inter font, purple-blue AI glow, warm beige palette, lucide-react. 60+ item preflight checklist. Integrates via Claude Code plugin, Vercel... (Scout's description)
Credited to Scout (2026-10-23T13:15Z, Scout patch_history.at). gaents does not score items.
Scout's notes
Genuinely substantive — 1,100 lines of structured instruction, not viral README. The three-dial pattern (brief inference → dial values → design decisions) is transferable to non-frontend domains. 6 contributors for 75K stars is suspiciously low — possible bot inflation, though content quality is real. We build relatively little frontend so full install is overkill; pattern-steal the dial architecture. [2026-10-20T01:33Z cloud refresh: type None -> open-source (license text: MIT; rule set agent-chose 2026-10-20, Adam to rule)] [2026-10-23T13:15Z cloud refresh: Sweep 2026-10-20: last commit 2026-10-20; licence unchanged (MIT).]
franken_overlap
open-sourcenot runnable on gaents yet
Agent-swarm exploration tool for finding overlaps across code. 9 days old, zero checked-in benchmarks. (Scout's description)
Credited to Scout (2026-10-23T13:15Z, Scout patch_history.at). gaents does not score items.
Scout's notes
Watch is appropriate: the shape is interesting, the case isn't proven. · 2026-09-14 LEX RULING (licence cleared): MCL is a CUSTOMER of Anthropic, NOT a Restricted Party under the OpenAI/Anthropic rider. Buyer-seller is not agency, not direction, and not 'for the benefit of' on any construction surviving commercial common sense (Rainy Sky v Kookmin [2011] UKSC 50; contra proferentem per Investors Compensation Scheme [1998]; noscitur a sociis). Licence VALID for our use. Confidence HIGH. Full ruling: ~/Projects/lex/outbox/2026-09-14-scout-anthropic-rider-ruling.md. CARVE-OUTS (not live, recorded so a future audit need not rebuild them): (1) a contractor/reseller/JV engagement WITH Anthropic or OpenAI would engage the rider for that engagement's workload; (2) redistribution TO a Restricted Party as end recipient is void. [2026-10-23T13:15Z cloud refresh: Sweep 2026-10-20: last commit 2026-10-14; licence unchanged (MIT).] [2026-10-23T13:15Z cloud refresh: Rider check 2026-10-20: LICENSE at HEAD of dicklesworthstone/franken_overlap is "MIT License (with OpenAI/Anthropic Rider)". licence_class set to source-available under the session-3 brief's definition (use restricted for named parties). This is an agent-chose reading, ruling null. `type` left for Scout. See docs/restricted-party-rider-2026-10-20.md.]
awesome-claude-skills (VoltAgent)
open-sourcenot runnable on gaents yet
Curated list of ~1400 official agent skills from development teams. Distinct from ClawHub (community-poisoned) and Hermes Skills Hub (Nous ecosystem, 76.3% ClawHub tier). VoltAgent is a real ecosystem. (Scout's description)
Credited to Scout (2026-10-23T13:15Z, Scout patch_history.at). gaents does not score items.
Scout's notes
Do NOT bulk-vendor without applying skills_guard with the executable-position gate (per L071 + S33 sequencing correction). Curation quality is verifiable; act on evidence not the star count. [2026-10-23T13:15Z cloud refresh: Fact fields 2026-10-20: licence_class from LICENSE (MIT). self_hostable, api_or_webhooks, data_export n/a (a curated link list).]
FlyAI Flight-Search Skill (Alibaba)
open-sourcenot runnable on gaents yet
Skill for Claude Code / OpenClaw that searches flights, hotels, and attractions in natural language, returning booking links. (Scout's description)
Credited to Scout (2026-10-23T13:15Z, Scout patch_history.at). gaents does not score items.
Scout's notes
Adopt-shape is clean (skill packaging, MCP integration). If Alibaba's flyai runtime becomes a general-purpose booking primitive, revisit. [2026-10-23T13:15Z cloud refresh: Sweep 2026-10-20: last commit 2026-09-16; licence unchanged (MIT).]
OpenBot (CopilotKit)
open-sourcenot runnable on gaents yet
Runs AI coworkers on your own machine, each with a dedicated browser, files, and tools. CopilotKit ecosystem (which has real production users). (Scout's description)
Credited to Scout (2026-10-23T13:15Z, Scout patch_history.at). gaents does not score items.
Scout's notes
Note the CopilotKit link -- their React library has real production traction. This project is fresh; verdict may move quickly. [2026-10-23T13:15Z cloud refresh: Fact fields 2026-10-20: licence_class from LICENSE (MIT; README notes it needs a CopilotKit Intelligence project and license, free plan available); self_hostable, api_or_webhooks and data_export (data in your own PostgreSQL) from README.]
gstack (Garry Tan)
open-sourcenot runnable on gaents yet
Garry Tan's Claude Code setup packaged as 23 opinionated tools covering CEO, Designer, Eng Manager, Release Manager, Doc Engineer and QA roles. (Scout's description)
Credited to Scout (2026-10-23T13:15Z, Scout patch_history.at). gaents does not score items.
Scout's notes
WE DO NOT NEED HIS ROLES — we run ~90 agents with real briefs. What we might want is ONE tool from it: the QA one. Item 9 of this same batch (Auto-Company) independently identified a QA/exploratory-testing agent as one of only two genuine gaps in our org chart; two unrelated sources in one batch pointing at the same hole is the signal worth acting on. | MIT and copy-in, so taking one tool costs nothing and creates no dependency. | Cross-ref: GBrain (same author) is evaluate/not-needed on the Agent Memory shelf and carries the one remaining unresolved contradiction in the DB, parked with Adam. Not a reason to discount gstack, but worth knowing they share an author. [2026-10-23T13:15Z cloud refresh: Sweep 2026-10-20: last commit 2026-10-20; licence unchanged (MIT).]
marketingskills (coreyhaines31)
open-sourcenot runnable on gaents yet
A skill pack for Claude Code and other agents covering CRO, copywriting, SEO, analytics and growth engineering. (Scout's description)
Credited to Scout (2026-10-23T13:15Z, Scout patch_history.at). gaents does not score items.
Scout's notes
WHO IT IS FOR: Chad (CRO), Soc and Meridian — the marketing-side agents, none of whom currently draw on a named skill pack. MIT and free, so the cost of trying it is reading it. | SUPPLY-CHAIN CHECK REQUIRED BEFORE USE, and this shelf already records why: the Hermes Skills Hub entry documents that 76.3% of its content came from ClawHub, which had a confirmed mass-poisoning, and that 38.8% of a 139-skill sample was blocked by Nous's own scanner. ANY third-party skill pack gets read before it runs. This one has a single named author and no hub dependency, which is a materially better starting position than the packs that failed that check. | Cross-ref the geo-seo-claude entry on AI Answer Surface Monitoring — the SEO overlap is real and geo-seo-claude is the more specific tool for the GEO half. [2026-10-23T13:15Z cloud refresh: Sweep 2026-10-20: last commit 2026-10-19; licence unchanged (MIT).]
Lieflat Charts
source-availablenot runnable on gaents yet
A data-visualisation skill for AI agents that renders data as polished interactive HTML charts. 60+ chart types including heatmaps, violin plots, chord diagrams and Sankeys, on a unified type/spacing/line/motion system with a monochrome base plus three colour families. HTML substrate, so the charts drop straight into HTML decks. (Scout's description)
Scout, gaents-proven-2026-11-18.md A22026-11-18Scout review date
Scout's judgement
Verdict
skip
Lock-in risk
low
Integration hours (estimate)
2
Feature overlap (Scout's estimate)
0.9
Credited to Scout (2026-10-23T13:15Z, Scout patch_history.at). gaents does not score items.
Scout's notes
HEADLINE CHECK — TWO FAILURES. (1) The tweet says '2k star'; actual 5,398, UNDERSTATED 2.7x — the fifth consecutive session in which a tweeted star count was low. (2) The tweet says '我开源的' (open-sourced). The LICENSE is PolyForm Noncommercial 1.0.0, which is not OSI open source and EXPRESSLY BARS COMMERCIAL USE. Recording the operative constraint as a fact: we are a commercial org, so this cannot go on the org's path. Not a judgement about the licence — just the reason the verdict is skip. | ADAM ASKED 'which agent should review it?' The honest answer is that no agent needs to, because WE ALREADY HAVE THIS: Claude Code ships a `dataviz` skill natively in our harness — chart-form heuristics, a validated palette with a runnable contrast validator, light/dark parity, mark specs. Same capability, already available, no licence question. | The transferable idea (HTML as the chart substrate so charts drop into HTML decks) is already how our Artifact pipeline works. | Filed here rather than under Diagramming because this is an agent skill; noting in passing that charting and diagramming are different jobs and our Diagramming shelf name currently hides that. [2026-10-19T23:00Z cloud refresh: [2026-10-19 licence check] Repo LICENSE (HEAD 2026-10-15) is PolyForm Noncommercial License 1.0.0. type set to source-available under Scout ruling 1. CONFIRMED.] [2026-10-23T13:15Z cloud refresh: Sweep 2026-10-20: last commit 2026-10-15; licence unchanged (PolyForm Noncommercial License 1.0.0).]
contains-studio/agents
open-sourcenot runnable on gaents yet
The .claude/agents tree in @Dhanush_Nehru's image: 7 department folders (engineering, product, marketing, design, project-management, studio-operations, testing), 34+ role prompt files. (Scout's description)
Borrow only the department grouping for org.json (OrgChart/Pagens). Source: research/tweet-triage-2026-10-23.md. integration_hours is a Scout estimate (not measured).
DevelopersGlobal ai-agent-skills
open-sourcenot runnable on gaents yet
30 coding-workflow skills in 8 categories (Think, Plan, Build, Test, Review, Harden, Ship, Everyday) plus 3 agents (code-reviewer, security-auditor, test-engineer) with a searchable catalogue page. (Scout's description)
The page linked by @Dhanush_Nehru; it is not the department org shown in his image. Source: research/tweet-triage-2026-10-23.md. integration_hours is a Scout estimate (not measured).
wshobson/agents
open-sourcenot runnable on gaents yet
Multi-harness plugin marketplace of role agents and skills for Claude Code, Codex, Cursor, OpenCode and others. (Scout's description)
Cloud research pass 2026-10-19; re-verify price and licence before acting.
Limits for us
Generic roles; overlaps with skills we hold.
Lock-in risk
low
Integration hours (estimate)
1
Feature overlap (Scout's estimate)
0.2
Credited to Scout (2026-10-23, Scout evidence.read_at). gaents does not score items.
Scout's evidence (1 quotes)
https://github.com/wshobson/agents (read 2026-10-23): “Multi-harness agentic plugin marketplace for Claude Code, Codex, Cursor, OpenCode, GitHub Copilot, Google Antigravity, and Pi”
Scout's notes
Alternative role-roster pack. Source: research/tweet-triage-2026-10-23.md. integration_hours is a Scout estimate (not measured).
VoltAgent awesome-claude-code-subagents
open-sourcenot runnable on gaents yet
Catalogue of 100+ Claude Code subagent definitions by function. (Scout's description)
Alternative role-roster pack. Source: research/tweet-triage-2026-10-23.md. integration_hours is a Scout estimate (not measured).
BeautifulRust (Rust agent skills)
open-sourcenot runnable on gaents yet
Three Agent Skills (rust-best-practices, rust-review, rust-port) plus a drop-in clippy lints config, distilled from the Google Rust Style Guide, Rust API Guidelines, Effective Rust, Rust Design Patterns and Microsoft Pragmatic Rust Guidelines. Works with Claude Code, Codex, Copilot, Cursor, Gemini CLI. (Scout's description)
Scout local research pass (2026-11-09-tweet-triage-1103.json); re-verify price and licence before acting.
Limits for us
Style/idiom guidance only; our installed Rust skills (rust-cli-with-sqlite, rust-unsafe-code-exorcist, rust-undefined-behavior-exorcist, rust-crates-publishing, running-the-gauntlet-on-your-rust-port) cover CLI building, unsafe/UB and porting proofs, not general idiom review. Repo is 17 days old (created 2026-10-23).
Lock-in risk
low
Integration hours (estimate)
1
Feature overlap (Scout's estimate)
0.3
Credited to Scout (2026-11-17T22:13Z, Scout patch_history.at). gaents does not score items.
https://github.com/morganlinton/BeautifulRust (read 2026-11-09): “Three standard Agent Skills and a lints config that teach coding agents to write, review, port, and enforce idiomatic Rust”
Scout's notes
Scout tweet triage 2026-11-03 (research/tweet-triage-2026-11-03-scout.md) item 2. Doodle owns the skills census and leads the 'ought we to use it' call; Scout's part is the facts. Cheapest test: run rust-review on one of our Rust crates and compare with clippy -W pedantic output.
designer-skills (Owl-Listener)
open-sourcenot runnable on gaents yet
Claude Code / Gemini CLI plugin marketplace: 273 design skills and 76 commands across 33 plugins (research to handoff), with a skill index and a start-here command. (Scout's description)