Executive Summary
Matt Wolfe's "9 Free AI Skills That Feel Like Cheat Codes" (June 2026) is a practical roundup of reusable instruction files that dramatically improve AI coding agent outputs — from virtual engineering teams to knowledge graphs to animation generation. For Steve's Hermes+OpenClaw workflow, several of these skills directly address pain points in the Provenance Labs pipeline: code review quality gating, AI-writing cleanup before vault publication, and knowledge retrieval across the growing second brain. Others (ReMotion, HyperFrames, front-end design) are less immediately relevant but represent capabilities that could be unlocked as the vault evolves toward richer media content.
Key Takeaways
- **Highest immediate value for Steve** — GStack, Stop Slop, and Graphify map directly to active pain points: production-grade code review before wrangler deploys, cleaner AI-generated brief text before vault publication, and a queryable memory layer over the Obsidian second brain.
- **Skills are the same mechanism Hermes already uses** — The video's "skill.md file as repeatable instruction" is exactly how Hermes Agent's own skill system works. Every skill Wolfe installs is compatible with Hermes Agent and can be installed by simply providing the GitHub URL.
- **Graphify's graph-as-memory approach complements session_search** — While Hermes has session_search for past conversation recall, Graphify builds structured, queryable knowledge graphs from codebases and note collections — a fundamentally different capability that could power vault content discovery.
- **Stop Slop directly improves brief quality** — The remaining AI-isms in auto-generated briefs (cliches, overused phrases, robotic construction) are exactly what Stop Slop targets. Integrating this step before vault deploy would produce cleaner, more professional-looking vault content.
- **Last 30 Days fills Steve's market intelligence gap** — Real-time sentiment research across Reddit, X, YouTube, Hacker News, and Polymarket would be valuable for trend-aware brief generation — particularly for health-system workforce development topics where public sentiment matters.
- **Front-end design skills are low priority currently** — The vault's Duke Dark theme is already established and stable. Taste and Front End Design skills would only matter if building new public-facing pages outside the vault system.
- **ReMotion and HyperFrames are future-facing** — Animation generation has no current application in the text-brief pipeline, but could enable video summaries or animated vault previews down the line.
Key Findings
1 GStack: Production-Grade Engineering Team for Hermes
GStack by Garry Tan (Y Combinator CEO) bundles 23 specialist roles and 8 power tools as slash-command skills — CEO, engineering manager, designer, code reviewer, QA lead, security officer, and release engineer. Each role is a self-contained skill.md file with system-level instructions for how to behave. For Steve, the GStack review and GStack QA skills are immediately applicable: before every wrangler deploy, running `/gstack-review` across staged changes would catch production bugs, missed edge cases, security issues, and UX problems that a single-pass Hermes review might miss. The GStack plan-CEO-review skill provides executive-level architecture validation on feature ideas — a natural fit before committing to complex Provenance Labs workflow changes. The install process is trivial: "install this for me" with the GitHub URL, exactly as shown in the video. [Source 1]
2 Stop Slop: Cleaner Vault Content
Stop Slop by Hardik Pandya is a single skill file that strips AI tells from generated text — removing cliches, overused phrases, and robotic language common in LLM output. For Steve's pipeline, this addresses a real quality gap: the briefs and slides currently deployed to the vault are AI-generated and occasionally carry the "delve into," "landscape," "tapestry of" patterns that Stop Slop targets. Integrating Stop Slop as a post-synthesis, pre-deploy step would give vault publications a more professional, human-authored feel. The skill is tiny (single file, fast install) and works with a simple prompt: "Use the Stop Slop skill to improve this content." [Source 1]
3 Graphify: Queryable Memory for the Second Brain
Graphify by Safi Shamsi converts codebases, docs, schemas, notes, and media into interactive, queryable knowledge graphs that AI agents can use as persistent memory. The visual graph output resembles Obsidian's built-in graph view, but the key differentiator is the queryable JSON layer — instead of the agent re-reading all markdown files every time (expensive on tokens), it queries the pre-built graph. For Steve's growing second brain in Obsidian, this means: "What are the biggest recurring themes in my notes?" or "What three vault-brief ideas are hiding in this knowledge base?" can be answered in a fraction of the tokens. Graphify also works on codebases — running it on the Provenance Labs skills directory or the state-repo would produce an instant dependency map. [Source 1]
4 Understand Anything: Codebase Onboarding Maps
Understand Anything by Egonex AI creates interactive, exploratory knowledge graphs focused on human-readable codebase understanding — "graphs that teach are greater than graphs that impress." Unlike Graphify's agent-facing memory layer, Understand Anything prioritizes human onboarding: clickable flow diagrams showing how API routes, services, data models, and UI layers interconnect. For Steve, this is valuable if Rook or another collaborator needs to understand the Provenance Labs deployment codebase or the vault site structure. A one-command run (`/understand`) produces a browsable HTML map that answers "where should a new developer start?" without digging through every file. [Source 1]
5 Last 30 Days: Real-Time Sentiment Research
Last 30 Days by M Van Horn performs grounded sentiment research across Reddit, X, YouTube, Hacker News, Polymarket, GitHub, and general web search — then synthesizes findings into a structured report with source counts. Wolfe demonstrates this on "Claude Fable," pulling 21 Reddit threads, 20 YouTube videos, 27 Hacker News stories, 19 GitHub items, and 4 Polymarket markets in a single run. For Steve's health system workforce development briefs, this skill could add a public-sentiment dimension: "What are people saying about AI in healthcare workforce training right now?" The skill also generates shareable HTML reports — directly publishable alongside vault briefs. [Source 1]
6 Front End Design and Taste Skills: UI Polish (Lower Priority)
Anthropic's Front End Design skill and Leonxlnx's Taste skill (47K+ GitHub stars) both aim to improve AI-generated UI aesthetics. The video's head-to-head test on FutureTools.io shows clear quality improvements over unaided model output — more colorful layouts, better spacing, refined typography. For Steve, these skills are low priority because the Provenance Labs vault design system (Duke Dark theme) is already established and stable. They would become relevant only when building new public-facing pages or dashboards outside the vault — or if Steve's standards repo includes UI templates that need aesthetic baselines. [Source 1]
7 ReMotion vs. HyperFrames: Animation Generation (Future Use)
ReMotion (Remotion Dev) and HyperFrames (HeyGen) both generate programmed motion graphics from a single text prompt — iPhone text conversations, logo reveals with particle explosions, animated stock charts. HyperFrames generally produces more polished results in Wolfe's tests (better iPhone mockup, better Nvidia stock chart animation). For Steve's current workflow, animation has no application in the text-brief-to-vault pipeline. However, if the vault evolves to include video summaries, animated preview cards, or narrated brief-to-podcast companion visuals, HyperFrames would be the recommended starting point. Wolfe notes that for truly polished animations, a human After Effects artist is still superior — these tools handle "basic graphics" well. [Source 1]
Risks, Gaps & Uncertainty
- These briefs are derived from video content, not peer-reviewed research or hands-on testing by the author
- Speaker perspectives (Matt Wolfe) reflect a content-creator and "vibe coder" perspective, not enterprise software engineering standards
- Transcripts may be auto-generated and contain terminology errors or omissions from the live demonstrations
- GStack's 23 specialist roles overlap significantly with Hermes' built-in capabilities — installing it may cause role conflicts or duplicate behaviors that need deduplication
- Graphify's knowledge graph query layer depends on consistent markdown formatting in the Obsidian vault — fragmented or unstructured notes would produce sparse graphs
- Last 30 Days relies on real-time web scraping which may face rate limits, API changes, or bot detection from target platforms
- None of the skills have been specifically tested with Hermes Agent — the video claims "works with Hermes" but provides no Hermes-specific demonstrations
- The animation skills (ReMotion, HyperFrames) require Node.js runtime and significant disk space for FFmpeg/codec dependencies — suitability for this containerized VPS environment is unverified
Recommended Next Actions
Install GStack in Hermes — Run `install this for me https://github.com/garrytan/gstack` in a Hermes session. Run `/gstack-review` on the current vault site directory before the next wrangler deploy to validate the review quality.
Patch Stop Slop into the deploy pipeline — After brief synthesis and before HTML generation, add a Stop Slop pass on the brief content. Test on an existing published brief first.
Test Graphify on the Obsidian vault — Run Graphify on the second brain wikis/journals directory. Verify the queryable JSON layer produces accurate theme detection before integrating into the pipeline.
Install Last 30 Days for one research brief — Use it on a health-system workforce development topic to evaluate sentiment-report quality. If good, add as an optional pre-brief research step.
Benchmark Understand Anything on the state-repo — Run it on `/shared-workspace/state-repo/` to produce a browsable onboarding map for future collaborators.
Defer animation and design skills — Tag ReMotion, HyperFrames, Front End Design, and Taste as "evaluate later" — revisit if the vault expands to include video or public-facing dashboards.
Create a `ai-skills-workflow` Hermes skill — Once the best candidates are validated, create a skill documenting which external skills are installed, how they're integrated into the pipeline, and any deduplication with Hermes' built-in capabilities.
Annotated References
[1] Matt Wolfe. (2026). 9 Free AI Skills That Feel Like Cheat Codes. YouTube. https://www.youtube.com/watch?v=STH929HARLo
Primary source. Showcases 9 free AI coding skills/plugins with live installations and head-to-head comparisons, demonstrating cross-platform compatibility with Hermes, Codex, Claude Code, and other agent harnesses. Video transcript was used for brief synthesis.