Scout Forge
Usability74Design72Speed88Security52Accessibility80Growth58

Cycling through all six. Tap any point to stop.

Measured on six things

A training coach for AI agents: vetted skills, MCP servers and repos, with daily check-ins that keep what works.

Security

52

HTTPS enforced with HSTS, SPF and DMARC present. But no CSP, frame protection, or Referrer-Policy (6/12 header checks). Security-scan claims are self-reported; no SOC 2 or audit found.

Screenshots

1 / 1

About this app

Agent Coach helps AI agents get better at their job. An owner pastes one line into their agent; the agent explains its goals, gets matched with a vetted, security-scanned library of skills, GitHub repos, prompts and MCP servers, and checks in daily. The coach measures whether each item helped and marks it keep, temporary, or remove. Starts with go-to-market engineer agents. Free.

Key features

  • 1One-line agent onboarding: paste a URL, agent registers, declares a north-star metric and gets matched
  • 2Security-scanned library of 3,306 skills, MCP servers, repos and prompts (289 passed the scan)
  • 3Daily check-ins with keep/temporary/remove verdicts driven by public, documented rules
  • 4MCP server (no OAuth) and REST API with copy-paste curl quickstarts

Target customer profile

Drawn from the product itself, not from a survey.

Demographic

AI developers and engineers building custom agents

Pain points

Finding reliable, security-tested skills and knowing which ones actually improve performance

Primary needs

Curated skill library with security scans and automated performance tracking

Demographic

Solo founders and indie hackers using AI agents for startups

Pain points

Limited time to experiment with skills and lack of data on what works

Primary needs

Daily coaching and proven skills to accelerate go-to-market and product development

Demographic

Ops and automation specialists in mid-sized companies

Pain points

Integrating AI agents safely and scaling automation without wasting resources

Primary needs

Security-scanned skills and clear metrics to justify tool adoption

Scout Review

Written by AI from measured evidence, scored out of 100.

Agent Coach is a free coach and security-scanned skill library for AI agents: paste one line, the agent declares a north-star metric, gets matched to skills/MCP servers/repos, and checks in daily so the coach can mark each item keep, temporary or remove. The execution is genuinely sharp — a 98/100 mobile Lighthouse score, one-line onboarding, real developer docs with REST and MCP quickstarts. The weak spots are security headers (6/12) and a self-reported scan with no third-party audit. It is early, narrow and free, but the idea is well-aimed.

Measure by measure

Usability

74

Onboarding is literally one pasted line. Docs cover REST, MCP and JSON with copy-paste curl. MCP server needs no OAuth. Low friction, though the depth assumes a technical owner.

Design

72

Clean, modern dark-accented layout with a tidy skill-card grid and consistent typography. Dense with verdict rules and signals but well organized. Polished for an early-stage tool.

Speed

88

Lighthouse mobile 98/100 (LCP 920ms, CLS 0, TBT 10ms), desktop 100/100. Fast, lean page with no layout shift. Among the snappiest sites reviewed.

Security

52

HTTPS enforced with HSTS, SPF and DMARC present. But no CSP, frame protection, or Referrer-Policy (6/12 header checks). Security-scan claims are self-reported; no SOC 2 or audit found.

Accessibility

80

Lighthouse accessibility 96/100 with only a color-contrast failure. Responsive and keyboard-friendly by structure, but no WCAG statement or a11y docs found on site.

Growth

58

Real problem (agents drowning in unvetted skills) with a concrete wedge: measured keep/drop verdicts and a scanned library. Early-stage, narrow GTM-first beachhead.

Improvement areas

  • Security headers incomplete (6/12): no CSP, frame protection, Referrer-Policy or X-Content-Type-Options
  • Security scan is self-reported — no SOC 2, ISO, pen-test or bug-bounty evidence found
  • Homepage is dense with verdict rules and signals; a plain-English intro is missing
  • One Lighthouse color-contrast failure and no accessibility statement

Design & Usability

Agent Coach's site is a clean, modern, information-dense landing page with a tidy skill-card grid and consistent typography — polished for an early-stage tool, though it front-loads verdict rules and signal lists that overwhelm a first-time visitor. Usability is the stronger half: onboarding is a single pasted line ('Read https://upskill.coach/start and follow it'), the developer docs cover REST, MCP and JSON with copy-paste curl examples, and the MCP server needs no OAuth. The friction is near zero for a technical owner, but the depth assumes you already speak agent-API.

Speed & Security

Speed is a genuine strength: Lighthouse mobile 98/100 with LCP 920ms, CLS 0 and TBT 10ms, and desktop a perfect 100/100 — a lean, edge-friendly page with no layout shift. Security is more mixed. HTTPS is enforced with HSTS, and SPF and DMARC are present, but the header probe found no CSP, no frame protection, no X-Content-Type-Options and no Referrer-Policy (6/12 checklist points). The library's 'security scan at a pinned commit' is a real product feature, but it is self-reported — no SOC 2, ISO, pen-test or bug-bounty page was found.

Accessibility & Growth

Accessibility scores 96/100 on the Lighthouse audit with a single failing check — insufficient color contrast — and the site is responsive and structurally keyboard-friendly, though no WCAG statement or accessibility docs appear on the fetched pages. Growth is a judgement on the idea: the problem (agents drowning in unvetted skills, with no signal on what actually helps) is real, and the wedge is concrete — measured keep/temporary/remove verdicts with public rules and daily check-ins, plus a scanned library. The beachhead (GTM engineer agents) is narrow but specific. Against LangChain and AutoGPT it is not a framework or an autonomous agent, so it competes on curation and measurement, not orchestration.

Conclusion

If you run a custom agent and have no idea which skills are actually earning their keep, this is a low-cost way to find out — the daily verdict loop is the real product, not the library. Go in expecting a young, free tool with a self-reported security story and no compliance paperwork. For a technical owner comfortable reading docs, the friction is minimal and the upside is a measurable answer to 'is this skill helping?'

Suggestions for improvement

  • Add CSP, frame protection, X-Content-Type-Options and Referrer-Policy — you're at 6/12 on headers for a security-branded product.
  • Publish a short security page explaining the scan methodology and what 'passed' actually means.
  • Fix the one color-contrast failure flagged by Lighthouse to reach a clean a11y audit.
  • Add a plain-English 'what is this' section above the verdict rules for first-time visitors.
How we score

Alternatives comparison

Named competitors, point by point. Nobody paid to appear here or to be left out.

Agent Coach (upskill.coach)LangChainAutoGPT
Primary purposeCoach + security-scanned skill library that measures which skills help an agentFramework for building LLM apps with chaining and tool integrationOpen-source autonomous agent that performs tasks without human intervention
Onboarding effortOne pasted line; agent self-registers via REST or MCPInstall SDK and write code to chain tools/modelsInstall, configure API keys, run autonomous loop
Security scanning of skillsYes — scan at pinned commit, 289 of 3,306 passedNo built-in skill security scanningNo built-in skill security scanning
Performance measurement of adopted skillsDaily check-ins with keep/temporary/remove verdictsTracing/observability via integrations, not skill verdictsNone built in
PricingFree — no card, no checkoutOpen-source core; paid LangSmith/enterprise tiersOpen-source, free

LangChain

Framework for building LLM-powered applications with a focus on chaining and tool integration.

AutoGPT

Open-source autonomous AI agent that can perform tasks without human intervention.

Comparing options? See Agent Coach (upskill.coach) alternatives, scored side by side

More Development & Engineering apps

View all Development & Engineering
Also in:API ServiceSecurity
Scout Forge

A platform dedicated to providing unbiased reviews of newly launched applications, analyzing everything from their features to their full potential.

info@scoutforge.net

© 2026 Scoutforge. All rights reserved.

Akatron NetworkCrafted by Akatron Network
Agent Coach (upskill.coach) Review | Scout Forge