# Theo Tests Matt Pocock's Skills Repo and Potato's Pstack — Theo (t3.gg) > [!question] Lingering Questions > - How much of the "unslop" effect is the skill's specific pattern list versus simply asking any model to self-critique before finalizing an answer? > - What happens to skill reliability as a personal or team "fleet" grows past dozens of skills — does trigger-collision between two plausibly-matching skills become a practical problem? > - How portable are Cursor-specific Pstack skills in practice once run inside Claude Code or Codex? > [!tip] Article Ideas > - "Your skill's description is a trigger, not a summary" — a short, practical piece on writing skill/CLAUDE.md descriptions for auto-invocation, using Theo's thumbnail analogy. #articleseed > - A worked example applying "unslop"-style self-audit-for-AI-patterns instructions to Courtney's own newsletter drafts and deep-read Paraphrase/Newsletter sections. > [!link] Cross Reference > - [[3. Resources/Permanent Notes/A Skills Description Decides Whether Its Used Independent of Instruction Quality]] — existing Permanent note already covers this article's H1 claim (description-as-trigger) from a different source (Oleg Efimov, "Navigating Claude Code — Skills That Actually Fire"); NOT duplicated here, linked instead. #xref > - [[3. Resources/Permanent Notes/Curate a Personal Skill Fleet Instead of Bulk-Installing]] — new Permanent note from this article. #xref > - [[3. Resources/Permanent Notes/A Skill Can Be Evaluated by Pasting Its Text Into a Session]] — new Permanent note from this article. #xref > - [[3. Resources/Permanent Notes/A Skills Repository Is the Client Library for an MCP Server]] — adjacent claim from a different source (Penpot AI Kit) about skills-as-distribution-unit; different angle (MCP client-library framing vs. personal curation), worth holding alongside. #xref > - [[3. Resources/MOCs/AI & the Future of Knowledge Work MOC]] — best-fit MOC; this is general AI-agent tooling, not WordPress content. #xref > [!quote] Vocabulary > - **Fleet** — Theo's term for a single, personally-curated, cross-machine repo of agent skills, synced rather than accumulated ad hoc. > - **Model-invoked skill** — a skill the agent can pull into context on its own, based on its description matching the current task. > - **User-invoked skill** — a skill with "disable model invocation" set, requiring explicit human invocation (e.g., a slash command). ## Bib capsule Theo (t3.gg), "Theo Tests Matt Pocock's Skills Repo and Potato's Pstack," YouTube. A hands-on comparison video: Theo installs and live-tests skills from Matt Pocock's ~200k-star skills repo and Lauren ("Potato")'s Pstack repo against his own projects, using Claude Opus 5, Claude Sonnet, and GPT-5.6/Codex as the agents. ## Author's thesis **Q1 (what's the video about):** Whether two prominent third-party AI-coding-agent skill libraries (Matt Pocock's and Pstack) are actually worth adopting, tested by live use rather than description-reading. **Q2 (how does it develop that):** Theo demos several skills against real work — unslop's before/after writing-style rewrite, the grilling skill's tree-structured interview run across three different model configurations on his own project docs, and Pstack's arena/blast-radius/show-me-your-work skills — then closes with uniform advice: read before installing, start small, maintain a personal curated set rather than bulk-adopting either library wholesale. ## Key arguments - A skill's description functions as a trigger condition for auto-invocation, not a summary of its contents — this determines whether it fires, independent of how good the instructions inside it are. - Skills don't need to be formally installed to be evaluated; copy-pasting the raw markdown into an agent session and issuing the trigger phrase reproduces the behavior, short of skills depending on external scripts. - Neither library should be adopted wholesale — the value is in selective, informed curation into a small personal or team set. ## Claim atoms ### Skill descriptions function as triggers, not summaries > [!quote] ID:rw_h1_trigger > It shouldn't be called a description. It should be called a trigger. **Where/why**: Mid-video, explaining how agents select which skills to load into context. Reframes a common misunderstanding: optimize the description for *when the agent should reach for this*, not for summarizing everything the skill does. **Tags**: #xref #aphorisms ### Unslop rewrite is judged directly against unrewritten output > [!quote] ID:rw_h2_unslop > God, I didn't realize how quickly I adjusted to unslopped text and now when I read the sloped text, it hurts me. **Where/why**: During the live side-by-side comparison of an "unslopped" Codex/Soul response versus a plain ChatGPT response to the same prompt. Strongest first-person evidence in the video that a single skill measurably changed the reviewer's own reading behavior. **Tags**: #xref ### A skill can be evaluated by copy-pasting its text, no install required > [!quote] ID:rw_h3_copypaste > This is the easiest way to test out a texton skill. Literally just copy paste the text. **Where/why**: Just before the live grilling demo, where Theo pastes the grilling skill's raw markdown into two separate agent sessions without formally installing it. Lowers the barrier to trialing third-party skills. **Tags**: #xref #addendum ### Grilling-style interviews surface real documentation gaps across models > [!quote] ID:rw_h4_grilling > Cloud MD is badly stale. The repo countries its own vzero rule and there are two competing execution models in the tree. **Where/why**: Opus 5's first-pass output from grilling Theo about his own "LakeBed" project's AGENTS.md/CLAUDE.md. Independently corroborated by Sonnet's low- and high-effort runs surfacing overlapping questions — a rough replication check within the video itself. **Tags**: #xref ## Permanent candidates - **Skill descriptions function as triggers, not summaries** — NOT promoted. Already exists as [[3. Resources/Permanent Notes/A Skills Description Decides Whether Its Used Independent of Instruction Quality]] (from a different source, Oleg Efimov's "Navigating Claude Code — Skills That Actually Fire"). Linked in Cross Reference above instead of duplicated. - **Curate a personal skill fleet instead of bulk-installing a library** — promoted to [[3. Resources/Permanent Notes/Curate a Personal Skill Fleet Instead of Bulk-Installing]]. Checked existing Permanents for "fleet," "curated skill," "personal fleet" — no duplicate found. - **A skill can be evaluated by copy-pasting its text into a session, no install required** — promoted to [[3. Resources/Permanent Notes/A Skill Can Be Evaluated by Pasting Its Text Into a Session]]. Checked existing Permanents for "copy-paste... skill," "test... skill... without install" — no duplicate found. Distinct from the fleet-curation claim (this one is about lowering the cost of *evaluating* a skill; fleet-curation is about what to do once you've decided to keep one). ## Processing checklist - [x] All claim atoms have header + verbatim quote + where/why - [x] Triage callouts populated for every tagged highlight - [x] Checked existing Permanents first — found direct duplicate for the primary (H1) claim, linked rather than duplicated; promoted two genuinely distinct secondary claims - [x] Connected into MOC — [[3. Resources/MOCs/AI & the Future of Knowledge Work MOC]] (not WordPress content; general AI-tooling MOC is the correct fit per task framing) ## Links - [[3. Resources/MOCs/AI & the Future of Knowledge Work MOC]] — general AI-agent-tooling MOC; correct home since this is not WordPress content - [[3. Resources/Permanent Notes/A Skills Description Decides Whether Its Used Independent of Instruction Quality]] — because evidence-for; existing Permanent note this article corroborates from a second, independent source - [[3. Resources/Permanent Notes/Curate a Personal Skill Fleet Instead of Bulk-Installing]] — because evidence-for; new Permanent note from this article - [[3. Resources/Permanent Notes/A Skill Can Be Evaluated by Pasting Its Text Into a Session]] — because evidence-for; new Permanent note from this article - [[3. Resources/Readwise/Articles/Theo Tests Matt Pocock's Skills Repo and Potato's Pstack]] — source document