Price terms as published
Each line is the vendor’s own wording, split from the price capture checked on 2026-09-22. Nothing here is rounded, converted or re-worded — the full sentences stay in the review below. Lines that carry only a price continue the plan named on the line above, in the vendor’s own order.
| # | Plan / term | Checked |
|---|---|---|
| 1 | Free (open source, MIT) | 2026-09-22 |
One-liner: “Skills for Real Engineers. Straight from my .agents directory.” — Matt Pocock’s working skill collection for coding agents, open-sourced under MIT and now the reference point of the personal-skill-pack wave.
Verified facts (GitHub API + repository tree, 2026-09-22)
| Attribute | Value | Source |
|---|---|---|
| GitHub stars | 267,007 (22,549 forks, 511 open issues+PRs) | mattpocock/skills via GitHub API, 2026-09-22 |
| License | MIT | GitHub API license field + repo LICENSE file |
| Language | Shell | GitHub API |
| Created | 2026-02-03 | GitHub API |
| Last push | 2026-09-18 | GitHub API |
| Tagline (repo description) | “Skills for Real Engineers. Straight from my .agents directory.” | GitHub API |
| Plugin manifest | .claude-plugin/plugin.json v1.2.3 declares 25 skills |
repo file, fetched 2026-09-22 |
| SKILL.md files in tree | 38 (engineering 18, in-progress 9, productivity 7, misc 4) | recursive git tree, 2026-09-22 |
| Install surface | skills.sh listing (skills.sh/mattpocock/skills) |
README badge |
Star trajectory (time-stamped, not static)
~143k around mid-June → ~216k (08-14, ledger snapshot) → ~260–265k (09-15/19 readings) → 267,007 on 09-22 (API, this page’s primary). Growth has decelerated from the summer vertical but the repo keeps adding thousands per month — treat any star figure you see elsewhere by its date.
What’s actually in it
Skill files: markdown instructions (plus helper scripts) that coding agents
discover and load for specific tasks. Pocock’s catalog is model-invoked
engineering habits — tdd, code-review, domain-modeling,
diagnosing-bugs, grilling, to-spec, wayfinder — which is why it reads
as “senior engineer habits in file form” rather than a prompt-arcana dump.
The design idea worth stealing is progressive disclosure: an agent loads a
skill’s body only when the task calls for it, instead of stuffing every
instruction into the context window. That pattern, not any individual skill,
is why the repo became a reference point.
The 25-vs-38 catch
Count what you actually get installed, not what the folder tree implies. The
.claude-plugin manifest exposes 25 skills; the repo contains 38
SKILL.md files. The delta is mostly the skills/in-progress/ folder (9 more
drafts — pr, retro, loop-me, …) plus odds and ends that stay dormant
until the manifest lists them. Feed-side testing in mid-September recorded an
even smaller active subset (~12) at that moment; the manifest grows over
releases. Pin the plugin version (1.2.3 as of 2026-09-22) if your workflow
depends on a specific set being live.
The other catches
- It’s one person’s workflow. “Real engineers” is branding; the pack encodes Pocock’s TypeScript-era conventions (his own product is Total TypeScript), not universal law.
- Skills steer, they don’t sandbox. Whatever agent runs them decides how literally instructions are followed.
- Star count ≠ maintenance depth — 511 open issues+PRs on a solo-maintainer repo is normal for this scale, but don’t expect issue-response SLAs.
Who should use it
Developers already on Claude Code / Codex / Hermes-style agents who want a battle-tested starting set of engineering-habit skills — and honestly, anyone calibrating what the agent-skill ecosystem has become in 2026.
Repo facts from GitHub API, recursive git tree, and .claude-plugin/plugin.json
fetched 2026-09-22; trajectory figures are time-stamped snapshots labeled by
source date; the ~12-active observation is labeled feed-side testing. No skills
were executed for this page; no hands-on claims are made.