Daily AI Implementation Scout Council

Today's ranked top 20

#1

build nowClaude (Axion)34 / 35

Claude Code: Subagent Fork Default, Cross-Session Mentions, GitLab Marketplace repo

The latest Claude Code release (circa Aug 25-26, 2026) makes subagent forking on by default so fork subagents inherit the full conversation and prompt cache, adds cross-session @mention so Claude can reach another live session by name via SendMessage, and extends plugin marketplace support to bare GitLab repo URLs including nested subgroups.

What it does for you: Subagent fork default is the most operationally significant change: any agent spawned with subagent_type fork now gets the full conversation context and prompt cache automatically, which cuts redundant context setup across every multi-agent Axion flow. The @mention feature lets Hermes bridge sessions communicate with each other directly by name, which is a building block for multi-agent orchestration without external message queues. GitLab plugin marketplace support means YY can host private skills in a GitLab repo and have Claude Code resolve them the same way as GitHub repos.

In practice: A well-packed release. The fork default and @mention together lay the foundation for the multi-agent conversation graph that the Claude Agent SDK has been building toward. The GitLab support closes the gap for teams standardized on GitLab.

For: Claude (Axion). Claude Code is the direct session runtime; these changes affect every Claude Code session and every agent spawned from it.

Security4

Quality5

Auditability5

Useful to you5

Useful to community5

Buildable now5

Hermes5

Verdict: build now. Fork default changes how every existing Axion agent subagent spawn behaves. Review all CLAUDE.md files that define subagent spawning to confirm inherited context is intended. Update any that expect isolated context. The @mention feature is immediately usable in multi-bridge Hermes setups.

Build #1 Claude Code Subagent Fork Default and Cross-Session Mentions: use the ai-implementation-build-intake skill to build this safely. Source: https://github.com/anthropics/claude-code/releases. Save canonical skill/agent under AXION\Skills and AXION\Agents. Copy

Source · github.com/anthropics/claude-code/releases shows two consecutive releases in the Aug 24-26 window. Subagent fork default and @mention are explicitly in the What's changed list. GitLab token families (9 families: glrt-, gloas-, glptt-, glagent-, glimt-, glsoat-, glcbt-, glft-, glffct-) now get secret redaction. Full redaction applies to glpat- and gldt- tokens.

#2

newbuild nowBoth runtimes32 / 35

Cloudflare Kitesurf: Agent-First Browser on Workers repo

Kitesurf is Cloudflare's stateless agent-first browser running inside V8 isolates on Workers. It uses 3 to 7 times less CPU and memory than Chromium for typical agentic tasks such as screenshots and HTML extraction. It supports the Chrome DevTools Protocol so existing Puppeteer and Playwright clients work without code changes. Free while in beta.

What it does for you: Any Axion task that currently calls Cloudflare Browser Run for web scraping, screenshot capture, or page interaction can switch to Kitesurf by appending browser=kitesurf to the endpoint URL. No infrastructure changes are needed. The lower CPU cost makes bursty agent workloads cheaper, and the stateless design fits the unattended overnight Hermes pattern where sessions should not persist between runs. The public playground at kitesurf.cloudflare.app lets you test without writing any code.

In practice: Clean engineering story: strip everything a human browser needs (tabs, themes, extensions, sync) and keep only what an agent needs (token efficiency, CDP, scale). The V8 isolate approach avoids the heavyweight VM startup that slows Chromium-based browser automation.

For: Both runtimes. A drop-in Browser Run endpoint change: any Python or Node script in Axion or Hermes that calls the Cloudflare Browser Run API can opt in by adding browser=kitesurf as a query parameter.

Security4

Quality5

Auditability5

Useful to you5

Useful to community5

Buildable now4

Hermes4

Verdict: build now. The browser=kitesurf param is a one-line change in any existing Cloudflare Browser Run call. Swap it into the web-scraping utility used by the Axion bridge and run a latency and cost comparison before the next nightly run. Free beta removes the cost risk.

Build #2 Cloudflare Kitesurf Agent Browser Integration: use the ai-implementation-build-intake skill to build this safely. Source: https://developers.cloudflare.com/changelog/post/2026-08-06-kitesurf/. Save canonical skill/agent under AXION\Skills and AXION\Agents. Copy

Source · Announced Aug 6, 2026 at developers.cloudflare.com. Benchmark: 3 to 7 times less CPU and memory than Chromium on screenshot and HTML extraction tasks. Passes 235,000 web platform tests. Chrome DevTools Protocol compatible. Free while in beta. Confirmed absent from the seen registry as of this scouting run.

#3

newbuild nowBoth runtimes33 / 35

Claude Skills API GA repo

Anthropic's Skills API and Files API are now generally available on the Claude Platform and through Microsoft Foundry. The Skills API lets developers define callable skills that any Claude application can invoke by name, enabling modular agentic behavior without repeating prompt scaffolding across callers.

What it does for you: The Skills API turns the AXION Skills folder into something callable from any Claude app or Managed Agents workflow: define a skill once via the API, then call it by name from Claude chat, Hermes scripts, or any SDK consumer. The Files API GA means persistent file references can now be passed across API calls without re-uploading. Both are also live on Microsoft Foundry for Azure-hosted deployments.

In practice: A structural platform upgrade. Skills as first-class API objects are the missing link between the local AXION Skills folder and programmatic API consumers. The Files API GA completes the durable asset pipeline alongside Projects and Memory.

For: Both runtimes. The Skills API applies to both Claude Code sessions (skills as callable commands) and Hermes agent calls (skills as structured tool definitions accessible via the API).

Security5

Quality4

Auditability5

Useful to you5

Useful to community5

Buildable now4

Hermes5

Verdict: build now. Register the top Axion skills (hook-writer, speak-marc, campaign-brief) via the Skills API so they are callable from Claude chat and any future Managed Agents workflow without copy-pasting prompts. This is a direct upgrade to the existing AXION Skills architecture.

Build #3 Claude Skills API GA Registration: use the ai-implementation-build-intake skill to build this safely. Source: https://docs.anthropic.com/en/docs/build-with-claude/skills-api. Save canonical skill/agent under AXION\Skills and AXION\Agents. Copy

Source · Announced in Anthropic's August 20, 2026 developer platform release notes. Skills API is GA on claude.ai and via the API. Files API is also GA. Microsoft Foundry availability confirmed in the same release note. Skills API and Files API were listed as previously beta-only features now promoted to GA.

#4

newbuild nowBoth runtimes31 / 35

Claude Managed Agents: web_search and web_fetch Domain Controls repo

Claude Managed Agents now accept allowed_domains and blocked_domains parameters on the web_search and web_fetch tools, letting you restrict which sites an agent can reach during a run.

What it does for you: Any Hermes agent that uses web_search can now be sandboxed to a specific domain allowlist, so the AI scout newsletter agent cannot reach phishing or data-exfil sites even if a prompt tries to redirect it. For the MCL content pipeline, constraining web_fetch to approved source domains gives a clean audit trail of what the agent actually read during a run.

In practice: An obvious security primitive that was missing. Managed Agents with unrestricted web access were always a policy risk; this closes it without requiring a custom proxy or separate firewall.

For: Both runtimes. Applies to any Managed Agents API call where the web_search or web_fetch tool is used, covering both Hermes agent scripts and Axion scheduled tasks.

Security5

Quality4

Auditability5

Useful to you4

Useful to community4

Buildable now5

Hermes4

Verdict: build now. Add allowed_domains to all Axion Managed Agents calls that use web_search or web_fetch. Start with the AI scout newsletter agent (restrict to GitHub, Anthropic, MCP, and known tech sources) and the MCL content pipeline. This is a one-line API parameter change with no breaking effect on existing calls that omit it.

Build #4 Claude Managed Agents Domain Controls: use the ai-implementation-build-intake skill to build this safely. Source: https://docs.anthropic.com/en/docs/build-with-claude/agents/managed-agents. Save canonical skill/agent under AXION\Skills and AXION\Agents. Copy

Source · Documented in the Anthropic developer platform Aug 20, 2026 release. allowed_domains and blocked_domains are string arrays accepted by both web_search and web_fetch tool definitions in Managed Agents API calls. This was not available before the Aug 20 release.

#5

newtest firstBoth runtimes30 / 35

Claude Computer Use and Browser Use: Now Generally Available repo

The computer use tool and the browser use tool are now generally available on the Claude Platform and through Microsoft Foundry. They were previously in beta. Updated versions are also coming to Google Cloud's Vertex AI.

What it does for you: GA status means a stable API contract: the computer_use and browser_use tool schemas are version-locked and will not change without a deprecation notice. Any Axion automation that was waiting for GA before committing to these tools can now proceed. Browser use is the cleaner alternative to raw Playwright for Claude-driven web tasks, and computer use enables screen-level automation for tools without an API.

In practice: Expected but welcome. The GA commitment makes it safe to build production Axion flows on top of computer_use. The Vertex AI rollout broadens options if YY considers a GCP deployment path for Hermes.

For: Both runtimes. Computer use and browser use are API-level tools available to both SDK callers and Managed Agents, including Hermes automation scripts.

Security4

Quality4

Auditability5

Useful to you4

Useful to community5

Buildable now4

Hermes4

Verdict: test first. Run the browser_use tool against the Axion web automation test suite before replacing existing Playwright-based scrapers. GA status removes the API-change risk but the tool behavior on dynamic single-page apps should be verified for the specific pages Axion uses.

Build #5 Claude Computer Use and Browser Use GA Integration: use the ai-implementation-build-intake skill to build this safely. Source: https://docs.anthropic.com/en/docs/computer-use. Save canonical skill/agent under AXION\Skills and AXION\Agents. Copy

Source · Listed in Anthropic's August 2026 platform release notes (Aug 20, 2026). Computer use and browser use confirmed GA on the Claude Platform as of August 2026. Microsoft Foundry availability confirmed. Vertex AI availability listed as coming soon in the same release note.

#6

newtest firstClaude (Axion)27 / 35

Claude Remote Sessions Beta: Server-Side Scheduled Tasks repo

Claude now supports remote computer use sessions that run on Anthropic's servers, with session state saved to your Claude account and synced across devices. Scheduled tasks in remote sessions run server-side, so the local device does not need to be online.

What it does for you: For YY's unattended Axion automations, this is the cloud alternative to the local Hermes bridge: Claude can run a scheduled computer use task entirely server-side, which means the overnight scan and the morning newsletter could run even if the Windows box is off. This does not replace the Hermes bridge today, but it is the converging path for scheduled Claude Code sessions without a local runner.

In practice: Early beta, but the direction is clear: Claude is becoming a full agent host, not just a model. Remote sessions plus server-side scheduled tasks covers the same feature set as the Axion bridge, offered as a managed service.

For: Claude (Axion). Remote sessions run on Anthropic's servers; they apply to Claude Code and Managed Agents workflows but do not change the Hermes bridge runtime directly.

Security4

Quality4

Auditability5

Useful to you4

Useful to community4

Buildable now3

Hermes3

Verdict: test first. Spin up one remote session for a low-stakes scheduled task (for example, the daily AI scout scan) and compare reliability and cost against the local Hermes bridge before expanding. Beta status means the session API may change.

Build #6 Claude Remote Sessions Beta Integration: use the ai-implementation-build-intake skill to build this safely. Source: https://docs.anthropic.com/en/docs/remote-sessions. Save canonical skill/agent under AXION\Skills and AXION\Agents. Copy

Source · Announced in Anthropic's August 2026 platform release notes. Remote sessions in beta as of August 2026. Files and session state synced to Claude account. Scheduled tasks confirmed to run server-side with no device required.

#7

test firstClaude (Axion)28 / 35

Senior Dev Unbundled Plugin (Updated Aug 25, 111k Stars) plugin

A Claude Code plugin with 10 opinionated skills that make the AI agent push back on its own code, verify assumptions, and attack weaknesses before declaring work done. Listed in the Anthropic-managed claude-plugins-official directory and updated Aug 25, 2026.

What it does for you: Installing this plugin adds a set of skills that activate Claude Code's self-critique behavior: the agent challenges its own output rather than shipping the first working version. For YY's tool builds and Axion script development, this adds a built-in second opinion without spawning a separate review agent. The 111,000 star count across the wider ecosystem signals broad adoption in engineering teams that share Claude Code setups.

In practice: The 111k star count is unusual and suggests viral adoption. The 10-skill structure is clean: each skill targets a specific failure mode (verify assumptions, attack own code, check for regressions) rather than being a generic review prompt.

For: Claude (Axion). This is a Claude Code plugin installed in the Claude Code session; it does not affect Hermes API calls.

Security3

Quality4

Auditability4

Useful to you5

Useful to community5

Buildable now5

Hermes2

Verdict: test first. Install in a non-production Claude Code session and run it against the next Axion tool build before enabling globally. The push-back behavior can slow agentic loops if not tuned to the project context.

Install: /plugin install senior-dev-unbundled@claude-plugins-official from your Claude Code session. Review skill definitions in the plugin manifest before enabling. Not confirmed in enabledPlugins: check settings.json before enabling to avoid duplicate activation. Copy

Source · Listed in anthropics/claude-plugins-official (Apache-2.0 license, 3,900 forks on the official directory repo). Updated Aug 25, 2026. Reported star count: 111,000 on the wider plugin ecosystem. Self-described as 10 opinionated skills for code verification and self-critique.

#8

newtest firstBoth runtimes30 / 35

Langfuse v4: 165x Faster, Slack Alerts, Deterministic Checks, Metrics API v2 repo

Langfuse v4 (released Aug 17, 2026) delivers up to 165 times faster performance at scale, a new Metrics API v2 for aggregating cost, tokens, volume, latency, and scores across observations, a filter search bar combining typed filters and metadata in a shareable query, and quality threshold alerts that notify via Slack, webhooks, or GitHub Actions. Deterministic Python or TypeScript checks can now run against live observations.

What it does for you: The Slack alert integration maps directly to the Axion bridge alert pattern: set a latency or cost threshold for any Hermes agent and get a Telegram-forwarded Slack ping when it breaks. The deterministic Python check runner lets YY write small evaluation functions and run them automatically against every agent trace, replacing the manual spot-check that currently gates each Hermes deployment. The 165x speed improvement makes real-time tracing practical for the high-frequency bridge runs.

In practice: This is the observability release that makes agent quality loops practical for a solo operator. Slack alerts plus deterministic Python checks cover the two missing pieces in YY's current Hermes monitoring: automated quality gates and low-latency alerting. Note: the prior seen-registry entry covers the overview announcement at langfuse.com/changelog; this item covers the specific feature breakdown at the dedicated v4 page.

For: Both runtimes. Langfuse is an observability layer added to any Python or Node.js agent via the OpenTelemetry SDK; it integrates with both Axion scripts and Hermes agent calls.

Security4

Quality5

Auditability5

Useful to you4

Useful to community5

Buildable now3

Hermes4

Verdict: test first. Wire the Hermes bridge to Langfuse v4 via OpenTelemetry and configure a cost-per-run alert before the next nightly deploy. Confirm the 165x speed claim holds on the bridge trace volume before enabling deterministic checks on every run.

Build #8 Langfuse v4 Hermes Bridge Integration: use the ai-implementation-build-intake skill to build this safely. Source: https://langfuse.com/changelog/2026-08-17-langfuse-v4. Save canonical skill/agent under AXION\Skills and AXION\Agents. Copy

Source · Released Aug 17, 2026, documented at langfuse.com/changelog/2026-08-17-langfuse-v4. Performance claim: up to 165 times faster. Metrics API v2 aggregates 5 dimensions: cost, tokens, volume, latency, scores. Alerts support 3 delivery channels: Slack, webhooks, GitHub Actions. Open-source under Apache-2.0. Cloud v3 endpoints deprecated Nov 16, 2026.

#9

newtest firstBoth runtimes27 / 35

MLflow 3.6.0: Multi-Turn Eval, Side-by-Side Trace Comparison, Claude Code Tracing repo

MLflow 3.6.0 adds multi-turn conversation evaluation so complex agent conversations can be assessed end-to-end, a side-by-side trace comparison view in the UI for regression analysis, native Claude Code tracing support, and a background trace upload daemon that writes to local disk and retries on failure so agents never block on the network.

What it does for you: The multi-turn eval lets YY score a full Hermes run (prompt, tool calls, responses) as a single unit rather than scoring each turn in isolation, which catches failures that only appear in conversation flow. The side-by-side trace comparison is the key tool for verifying that a Hermes bridge upgrade did not regress quality: run the same scenario on old and new versions and compare. The background daemon means enabling MLflow tracing on the overnight bridge does not add latency to the run.

In practice: MLflow has become the pragmatic choice for agent evaluation: open-source, Databricks-backed, deep tracing with replay, and now multi-turn eval. The Claude Code native tracing integration is a direct fit for the Axion toolchain.

For: Both runtimes. MLflow 3 integrates via the mlflow.genai module into any Python agent; compatible with both Axion scripts and Hermes agent wrappers.

Security4

Quality4

Auditability4

Useful to you4

Useful to community4

Buildable now3

Hermes4

Verdict: test first. Instrument one Hermes agent run with mlflow.genai tracing and run a multi-turn eval against the last 10 bridge runs before wiring it into the CI gate. The 3.6.0 point release has low integration risk.

Build #9 MLflow 3.6.0 Hermes Eval Integration: use the ai-implementation-build-intake skill to build this safely. Source: https://mlflow.org/releases. Save canonical skill/agent under AXION\Skills and AXION\Agents. Copy

Source · MLflow 3.6.0 release notes list multi-turn conversation eval, side-by-side trace comparison view, and Claude Code tracing. Background daemon retries on failure and replays traces from disk after unexpected session exit. Open-source Apache-2.0. Databricks-maintained. MLflow 3.15.1 is the current patch as of scouting date.

#10

newwatchBoth runtimes26 / 35

MCP New Roadmap: Agentic Messaging Primitives, Agent Identity, HTTP Transport repo

The MCP Core Maintainers published a new roadmap on Aug 22, 2026, covering the next specification release and beyond. Five priority areas: agentic messaging primitives (server-initiated events), HTTP-native transport unification and hardening, agent identity and enterprise-ready security, improved primitives, and improved SDK developer experience. Each area has assigned Core Maintainers and Working Groups.

What it does for you: Agent identity is the most actionable signal for Axion: the spec is adding a standard way for an MCP server to verify which agent is calling it, which is the missing piece for per-agent access control in the Hermes tool layer. Agentic messaging primitives (server-initiated events) will enable push-based notifications from MCP tools to agents, removing the polling pattern that currently drives several Axion scheduled checks. Track the agent identity Working Group output before wiring any auth-gated MCP server to Hermes.

In practice: Roadmap, not a release. But the agent identity priority area has matured from horizon to active work, which means spec drafts will arrive in weeks rather than quarters. The five-area structure is the clearest statement yet of where MCP is heading.

For: Both runtimes. The MCP spec and roadmap govern the wire protocol for every MCP server and client in Axion and Hermes; both runtimes are affected by spec direction.

Security5

Quality4

Auditability5

Useful to you3

Useful to community5

Buildable now1

Hermes3

Verdict: watch. No code to ship today; this is a strategic alignment read. Bookmark the agent identity Working Group output and review the agentic messaging primitives proposal before the next MCP server build in Axion.

Build #10 MCP Roadmap Tracking: use the ai-implementation-build-intake skill to build this safely. Source: https://blog.modelcontextprotocol.io/posts/mcp-roadmap/. Save canonical skill/agent under AXION\Skills and AXION\Agents. Copy

Source · Published Aug 22, 2026 at blog.modelcontextprotocol.io. Five named priority areas each with assigned Core Maintainers and Working Groups. The roadmap page at modelcontextprotocol.io/development/roadmap is the living reference. Previous roadmap linked for comparison. Confirmed new: URL not found in seen-registry as of this scouting run.

#11

watchBoth runtimes25 / 35

alirezarezvani/claude-skills: 380 Claude Code Skills Across 8 Domains repo

A GitHub repo containing 380 Claude Code skills including 30 or more agents, 70 or more custom commands, covering engineering, marketing, product, compliance, C-level advisory, research, business operations, finance, and daily productivity. Compatible with Claude Code, Codex, Gemini CLI, and Cursor.

What it does for you: A fast scan of this collection could surface skills in the compliance, business operations, and finance domains that fill gaps in the current AXION Skills folder, particularly for MCL coaching workflow automation. The 30 included agents are a useful comparison set against the current Axion agent library.

In practice: A large collection that covers unusual domains (C-level advisory, compliance) not well represented in the official skill ecosystem. Quality varies in community collections of this size; needs a targeted browse rather than a bulk install.

For: Both runtimes. The skills are language-agnostic markdown files installable in both Claude Code sessions and Hermes subagent contexts.

Security3

Quality3

Auditability4

Useful to you4

Useful to community4

Buildable now4

Hermes3

Verdict: watch. Browse the marketing, business operations, and coaching-adjacent folders for candidate skills to add to AXION. Do not bulk-install: read each candidate skill definition before adding to avoid conflicts with existing AXION CLAUDE.md rules.

Build #11 alirezarezvani claude-skills Selective Review and Install: use the ai-implementation-build-intake skill to build this safely. Source: https://github.com/alirezarezvani/claude-skills. Save canonical skill/agent under AXION\Skills and AXION\Agents. Copy

Source · github.com/alirezarezvani/claude-skills README lists 380 skills, 30 or more agents, 70 or more custom commands as of the August 2026 scouting date. Coverage spans 8 domains. Compatible with 8 coding agents including Claude Code, Codex, Gemini CLI, and Cursor.

#12

watchBoth runtimes22 / 35

rohitg00/awesome-claude-code-toolkit: 135 Agents, 176+ Plugins, 20 Hooks repo

A curated collection of Claude Code resources including 135 agents, 35 skills, 42 commands, 176 or more plugins, 20 hooks, 15 rules, 7 templates, 14 MCP configs, 26 companion apps, and 52 ecosystem entries.

What it does for you: The 14 MCP configs are the most actionable subset for Axion: pre-built MCP server configurations that can be dropped into CLAUDE.md. The 20 hooks are worth reviewing against the current Axion hooks setup to see if any patterns are missing. The 135 agents list is a useful map of what the community is building.

In practice: Larger collection than alirezarezvani. The hook and MCP config sections carry the most signal; the 135 agents and 176 plugins are likely noisy without a quality filter. Worth a targeted browse rather than a full read.

For: Both runtimes. The collection includes agents, skills, plugins, MCP configs, and hooks usable across both Claude Code and Hermes contexts.

Security3

Quality3

Auditability4

Useful to you3

Useful to community4

Buildable now3

Hermes2

Verdict: watch. Scan the hooks and MCP configs sections only. Compare the 20 hooks against the existing Axion hooks setup and adopt any that cover gaps. Do not install the plugin set wholesale.

Build #12 rohitg00 Claude Code Toolkit Hook and MCP Config Review: use the ai-implementation-build-intake skill to build this safely. Source: https://github.com/rohitg00/awesome-claude-code-toolkit. Save canonical skill/agent under AXION\Skills and AXION\Agents. Copy

Source · github.com/rohitg00/awesome-claude-code-toolkit README lists: 135 agents, 35 skills, 42 commands, 176 or more plugins, 20 hooks, 15 rules, 7 templates, 14 MCP configs, 26 companion apps, 52 ecosystem entries. Confirmed active as of August 2026 scouting date.

#13

newwatchStandalone tool25 / 35

Claude Admin API GA: Enterprise User Management Endpoints repo

The Admin API user-management endpoints for Claude Enterprise organizations are now generally available. Endpoints cover members, invites, groups, and custom roles.

What it does for you: If YY or MCL moves from Pro to an Enterprise Claude account, the Admin API lets you automate onboarding and off-boarding via the API rather than the dashboard. For the current individual Pro plan, this is a future reference rather than an immediate action. It is worth noting for MCL team expansion planning.

In practice: Completeness feature for Claude Enterprise teams. Not relevant to YY's current individual Pro plan but a useful data point for MCL team growth planning.

For: Standalone tool. Admin API user management is an enterprise organization control plane feature; it does not run inside Axion or Hermes agent sessions.

Security5

Quality4

Auditability5

Useful to you2

Useful to community4

Buildable now3

Hermes2

Verdict: watch. File as a reference for when MCL adds team members to a shared Claude account. No immediate action needed on the current Pro plan.

Build #13 Claude Admin API Integration Planning: use the ai-implementation-build-intake skill to build this safely. Source: https://docs.anthropic.com/en/docs/administration/administration-api. Save canonical skill/agent under AXION\Skills and AXION\Agents. Copy

Source · Announced in Anthropic's August 20, 2026 developer platform release notes. Four endpoint groups are GA: members, invites, groups, custom roles. Requires a Claude Enterprise organization subscription.

#14

newwatchStandalone tool25 / 35

Claude Academy: Anthropic Learning Hub with Courses, Badges, and Personalized Paths repo

Anthropic launched Claude Academy, a new learning hub offering courses, tutorials, badges, and personalized learning paths for safe and effective AI use. Available to all Claude users.

What it does for you: The badge and course structure could serve as a credential signal for MCL clients asking about AI literacy. For YY personally, the personalized recommendations might surface gaps in advanced Claude API usage patterns. The resource is most useful for onboarding MCL team members or coaching clients to structured AI usage with an Anthropic-backed credential.

In practice: Expected at Anthropic's scale. The badge system is the most commercially interesting piece: if Anthropic Academy badges gain market recognition the way Google Cloud certifications did, they become a training product signal for MCL.

For: Standalone tool. Claude Academy is a learning website; it is not integrated into any Axion or Hermes runtime.

Security5

Quality3

Auditability5

Useful to you2

Useful to community5

Buildable now4

Hermes1

Verdict: watch. Bookmark and share with MCL team for client education context. Review the advanced developer courses to identify any Anthropic-endorsed patterns that differ from current Axion practices.

Build #14 Claude Academy Evaluation and MCL Positioning: use the ai-implementation-build-intake skill to build this safely. Source: https://anthropic.com/academy. Save canonical skill/agent under AXION\Skills and AXION\Agents. Copy

Source · Announced in Anthropic's August 2026 platform release notes. Offers courses, tutorials, and badges. Personalized recommendations included. Free to all Claude users.

#15

watchClaude (Axion)21 / 35

composio-community/awesome-claude-plugins: Curated Plugin List from Composio plugin

Composio's community-maintained curated list of Claude Code plugins covering custom commands, agents, hooks, and MCP servers available through the Claude Code plugin system.

What it does for you: A fast way to survey the plugin ecosystem for Axion-relevant entries: Composio's list focuses on integration and automation plugins, which are more directly relevant to YY's workflow than general coding tools. Browse the MCP server plugins section for any connectors that overlap with Axion's current tool gaps.

In practice: Composio is a known integration platform in the agent ecosystem; their curation lens is more business-process-focused than developer-tool-focused, which is a useful complementary angle to the official Anthropic directory.

For: Claude (Axion). Claude Code plugins apply to Claude Code sessions only; they do not change Hermes API call behavior.

Security3

Quality3

Auditability4

Useful to you3

Useful to community3

Buildable now3

Hermes2

Verdict: watch. Browse once for MCP server plugins relevant to MCL operations (CRM, scheduling, content tools). Do not install from this list without reading each plugin's manifest first.

Install: browse composio-community/awesome-claude-plugins on GitHub, then run /plugin install <plugin-name>@<source> for each entry you select. Review each plugin manifest before enabling. Copy

Source · github.com/composio-community/awesome-claude-plugins confirmed active as of August 2026. Composio operates composio.dev, a commercial agent integration platform. Community list with no formal review cadence published.

#16

watchBoth runtimes23 / 35

Microsoft Agent Framework Declarative Package: Beta to Release Candidate repo

The agent-framework-declarative package in the Microsoft Agent Framework has moved from beta to release candidate status, stabilizing the declarative agent definition syntax for Python and promoting it from beta in the Microsoft Agent Framework 1.0 GA ecosystem.

What it does for you: The declarative syntax lets you define agent behavior as structured Python objects rather than procedural code, which is closer to the CLAUDE.md pattern already used in Axion. Reaching RC status means the API is stable enough to build against without planning for a major migration in the near term.

In practice: Incremental maturation of the Microsoft Agent Framework ecosystem. The RC designation is meaningful: it signals that Microsoft is committing to backward compatibility for this package going into the 1.0 cycle. Note: the MAF Harness (CodeAct + Hyperlight) was covered on July 31; this item is specifically about the declarative package promotion.

For: Both runtimes. The MAF declarative package works in any Python environment including Axion scripts and Hermes agent wrappers.

Security4

Quality3

Auditability4

Useful to you3

Useful to community3

Buildable now3

Hermes3

Verdict: watch. Review the declarative syntax against the current CLAUDE.md agent definition pattern in Axion. If the declarative format offers a cleaner way to define Hermes agents than the current procedural approach, prototype one agent before the next milestone.

Build #16 MAF Declarative RC Prototype: use the ai-implementation-build-intake skill to build this safely. Source: https://learn.microsoft.com/en-us/agent-framework/support/upgrade/python-2026-significant-changes. Save canonical skill/agent under AXION\Skills and AXION\Agents. Copy

Source · Listed in the Microsoft Python 2026 significant changes guide. agent-framework-declarative moved from beta to RC status in August 2026. Part of the Microsoft Agent Framework 1.0 GA ecosystem (GA April 3, 2026). Supports Anthropic as a non-Microsoft provider alongside OpenAI, Bedrock, Gemini, and Ollama.

#17

newwatchBoth runtimes20 / 35

KTX Skills Kit: Executable Context Layer for AI Agent Data and Analytics repo

KTX launched its Skills Kit on Aug 18, 2026, connecting AI agents in Claude, ChatGPT, Codex, and Cursor to market data, portfolio management, order execution, and prediction markets. The kit provides an executable context layer for data and analytics agents querying analytical databases.

What it does for you: The architectural pattern is directly transferable: KTX demonstrates how to wrap a business data source (market data) as a skill callable from any agent. The same pattern could be used in Axion to wrap MCL CRM data, GHL event data, or Perf Collector metrics as agent-callable skills. The crypto trading execution is not a direct MCL fit, but the skill kit pattern is.

In practice: The execution layer is crypto-specific. The data access and analytical query pattern is generic and worth studying for the Axion data skill architecture. KTX included because of the pattern value, not the use case.

For: Both runtimes. KTX Skills Kit exposes data and execution as callable tool definitions that any agent runtime can call; compatible with both Claude-based and standalone execution.

Security3

Quality3

Auditability4

Useful to you2

Useful to community3

Buildable now3

Hermes2

Verdict: watch. Read the ktx repo README for the skill definition and executable context pattern rather than the specific trading use case. Reference design for the Axion data-access skill layer.

Build #17 KTX Skills Kit Pattern Study: use the ai-implementation-build-intake skill to build this safely. Source: https://github.com/Kaelio/ktx. Save canonical skill/agent under AXION\Skills and AXION\Agents. Copy

Source · Launched Aug 18, 2026 in Seoul. github.com/Kaelio/ktx repo describes the skill kit as an executable context layer for data and analytics agents. Compatible with Claude Code, Codex, ChatGPT, and Cursor. Decrypt.co coverage confirms the launch date and feature set.

#18

newwatchBoth runtimes24 / 35

awesome-claude-code: 13 New Agent Skills and Resources Added Aug 23 skill

The hesreallyhim/awesome-claude-code hand-curated list added 13 agent skills and related Claude Code resources on Aug 23, 2026. A separate batch of 9 resources was also added on Aug 20.

What it does for you: The hesreallyhim list is the most selective community skill list: 13 additions in one commit is a meaningful batch from a list that curates carefully. Scan the Aug 23 additions for any skill in YY's gap areas (evals, data access, MCL content automation) before they proliferate elsewhere.

In practice: This list has been consistently ahead of the broader ecosystem in discovering useful skills before they go viral. The Aug 23 batch is worth a direct read of the commit diff to see the specific 13 items.

For: Both runtimes. Agent skills in this list are markdown-format Claude Code skills compatible with both Claude Code sessions and Hermes subagent contexts.

Security4

Quality3

Auditability4

Useful to you3

Useful to community4

Buildable now3

Hermes3

Verdict: watch. Read the Aug 23 commit diff on the repo to see the specific 13 additions. Shortlist any that map to AXION gaps and run them through the AXION skill gate (test in Claude Code session before registering in AXION).

Build #18 awesome-claude-code Aug 23 Batch Review: use the ai-implementation-build-intake skill to build this safely. Source: https://github.com/hesreallyhim/awesome-claude-code/commits/main. Save canonical skill/agent under AXION\Skills and AXION\Agents. Copy

Source · 13 agent skills and related Claude Code resources added in a single commit on Aug 23, 2026. A separate batch of 9 resources added Aug 20, 2026. Repository described as hand-picked rather than automatically populated. The list description says it covers skills, agents, status lines, developer tooling, and plugins.

#19

newwatchBoth runtimes23 / 35

Google ADK 2.0 Stable: Python, TypeScript, Java, Go with A2A and Visual Agent Designer repo

Google ADK 2.0 has reached stable status across all four language SDKs: Python, TypeScript, Java, and Go. Key additions include native A2A (Agent-to-Agent) protocol integration at v0.2, a visual Agent Designer in the Google Cloud console, standardized structured logging with slog, and graceful OS signal handling for container termination. Note: ADK Go 2.0 was covered in this newsletter on July 10; this item covers the Python stable release and overall 2.0 GA.

What it does for you: The A2A protocol integration in ADK 2.0 is a cross-framework signal: A2A allows Claude-based agents and Google ADK agents to delegate tasks to each other, relevant if MCL clients ever run a mixed-vendor agent setup. The visual Agent Designer is a useful comparison for thinking about Axion agent design tooling. Stable status makes ADK 2.0 the reference implementation for multi-framework agent interoperability patterns.

In practice: The Python stable release is the most relevant for YY given the Axion Python stack. The A2A integration is the architectural differentiator; it signals Google's commitment to cross-vendor agent communication standards. The separation from the OpenAI ecosystem noted in the DEV Community post is strategically relevant.

For: Both runtimes. Google ADK is a cross-runtime Python framework; agents built with it can run alongside or as alternatives to Axion's current Python stack.

Security4

Quality4

Auditability4

Useful to you3

Useful to community4

Buildable now2

Hermes2

Verdict: watch. Monitor the A2A protocol implementation in ADK 2.0 as it is relevant to any Axion agent that may need to communicate with non-Claude agents in MCL client environments. No immediate build action; file as a strategic reference.

Build #19 Google ADK 2.0 A2A Pattern Study: use the ai-implementation-build-intake skill to build this safely. Source: https://dev.to/peytongreen_dev/google-adk-20-is-stable-why-that-makes-the-openai-split-matter-more-3c6l. Save canonical skill/agent under AXION\Skills and AXION\Agents. Copy

Source · ADK 2.0 stable confirmed across Python, TypeScript, Java, and Go SDKs. ADK Go 2.0 tagged v2.1.0 released June 30, 2026. Python SDK upgraded to Google GenAI SDK v1.65.0. A2A integration via github.com/a2aproject/a2a-go v0.3.15. DEV Community post confirms stable designation. ADK Go 2.0 was previously covered July 10 in this newsletter under a different URL.

#20

watchStandalone tool21 / 35

quemsah/awesome-claude-plugins: Plugin Adoption Metrics via n8n repo

An automated collection of Claude Code plugin adoption metrics across GitHub repositories using n8n workflows, tracking which plugins are gaining traction in the wider ecosystem.

What it does for you: Rather than manually searching GitHub stars and install counts, this workflow automates the collection of plugin adoption signals. YY can run it as a weekly check to see which Claude Code plugins crossed thresholds since the last scout run, without spending time manually tracking the official directory. The n8n design means it can be added to the Axion automation stack without custom code.

In practice: A practical meta-tool for staying current on the plugin ecosystem. It operationalizes the plugin scout lens that this newsletter runs manually each day.

For: Standalone tool. This is a standalone n8n workflow for research and tracking; it is not integrated into Claude Code or Hermes directly.

Security3

Quality3

Auditability4

Useful to you3

Useful to community3

Buildable now3

Hermes2

Verdict: watch. Review the n8n workflow for feasibility in the Axion automation stack. If it integrates cleanly with the existing n8n setup, run it weekly as a supplement to the daily scout.

Build #20 quemsah awesome-claude-plugins Tracker Evaluation: use the ai-implementation-build-intake skill to build this safely. Source: https://github.com/quemsah/awesome-claude-plugins. Save canonical skill/agent under AXION\Skills and AXION\Agents. Copy

Source · github.com/quemsah/awesome-claude-plugins describes an n8n automation workflow for tracking Claude Code plugin adoption metrics across GitHub repositories. Confirmed active as of August 2026 scouting date.