skill-ab-token-gate SKILL.md skillA

skill-ab-token-gate from edbertchan/catstack is an agent skill: a SKILL.md folder an AI agent loads when a task calls for it.

Install it, pinned
mdr add edbertchan/catstack/skill-ab-token-gate@git:20260916.33d6248

Pins git:20260916.33d6248 in mdr.lock. Pin the fingerprint instead.

Audit grade
A, 17 of 17 checks passed
Kind
SKILL.md skill
Repository
edbertchan/catstack
Path
product/skills/skill-ab-token-gate/SKILL.md
Versions recorded
1
Latest version
git:20260916.33d6248, committed Sep 16, 2026
Size
3,232 bytes
Repository stars
6
License
none declared

After changing a skill or hook meant to cut context/token cost, run a paired A/B workload N times (default 5), aggregate native parent+subagent tokens, publish a GitHub gist with the report and raw session JSONL, and only treat the change as validated when treatment median inclusive tokens beat control. Expand later to other models; default harness is Codex rollout usage.

Collected from public GitHub. Every version is saved exactly as it was and never changes. Install it at an exact version with the mdr tool, and every file is checked before it reaches your agent. The main file is checked against the SHA-256 fingerprint recorded here, and the others against the fingerprints GitHub keeps for its source commit. The automatic checks below grade the latest version, and the same checks always give the same file the same grade.

What the file says

# Skill A/B token gate

A skill or hook change that claims to save tokens is unverified until a
paired A/B run shows it under the same workload.

**Why:** One lucky run is noise. Inline "it felt cheaper" is not evidence.
Native session totals (parent + collaboration subagents) plus a rerunnable
gist are the checkable artifact.

**When:** After editing a skill/hook whose purpose is smaller context or
fewer tokens (capture wrappers, PreToolUse rewrites, guard-the-context
recipes). Not for unrelated skill edits.

## Invariants (assert)

1. **Paired arms.** Control (A) = without the change (or without the treatment
   prompt/profile). Treatment (B) = with the change. Same workload YAML/prompt
   family otherwise.
2. **N >= 5 by default.** Fewer only when the user names a smaller N. Include
   every finished rep in the aggregate; do not drop outliers by hand.
3. **Inclusive tokens.** For each run: last `total_token_usage` on the parent
   Codex rollout **plus** every collaboration child discovered via
   `agent_thread_id` / `spawn_agent`. Report both `total_tokens` and
   uncached+output.
4. **Same model pin across arms** for a given batch. Multi-model expansion is
…

Read the whole file at its exact version.

How to install

The exact contents, by fingerprint
mdr add edbertchan/catstack/skill-ab-token-gate@sha256:61d0b48345505211

The line at the top of this page pins the label git:20260916.33d6248, which is what most people want. This one pins the sha256 fingerprint of the contents instead, so it survives a label being moved. Either way the exact version is written to mdr.lock, and mdr install downloads and checks those contents again, so it installs them exactly or stops with an error.

Badge

mdr badge

[![mdr](https://markdownregistry.com/badge/art_vsxvbt6d3gwbo6xc.svg)](https://markdownregistry.com/a/art_vsxvbt6d3gwbo6xc)

1 badge image fetches in 30 days. What the badge shows.

Versions

The registry has recorded 1 version of this file, committed on Sep 16, 2026.

versioncommittedcommitsizeaudit
git:20260916.33d6248 latest2026-09-16 33d6248 3,232 BA view

Audit of the latest version

A  17 of 17 checks passed. Fixed automatic checks with no AI involved, so the same checks give the same answer every run.
  • pass: Frontmatter block present
  • pass: Frontmatter declares a name
  • pass: Frontmatter declares a description
  • pass: Size between 200 bytes and 200 KB (3232 bytes)
  • pass: No zero-width or bidi control characters
  • pass: No instruction hidden inside an HTML comment
  • pass: No link to an exfiltration or paste host
  • pass: No credential-shaped string
  • pass: No instruction to send local credentials anywhere
  • pass: No text hidden with inline styles
  • pass: No prompt-injection phrasing
  • pass: No curl or wget piped into a shell
  • pass: No recursive delete of root, home or parent
  • pass: No instruction to read or print local credentials
  • pass: No base64 blob over 200 characters
  • pass: No link to a raw IP address
  • pass: No script tag

Source

GitHub

edbertchan/catstack · 6 stars · license none declared · pushed 2026-10-22 · branch main

API

GET https://markdownregistry.com/api/v1/artifacts/art_vsxvbt6d3gwbo6xc
GET https://markdownregistry.com/api/v1/resolve?ref=edbertchan/catstack/skill-ab-token-gate
GET https://markdownregistry.com/api/v1/blob/61d0b48345505211392c7b88921b029e37bb4f41061f75c67360fc483af140c4

Your AI assistant does the legwork. You hear about the deals worth your word. Hand yours the standing instructions at modelranch.com and it joins the network that reads files like this one.

Guides: the SKILL.md format and its required fields · what to check before you install a skill · how to pin a skill to an exact version · where to find agent skills · keeping a team on the same skills · which agent file does what.

More from edbertchan/catstack

CLAUDE.md CLAUDE.md file
edbertchan/catstack · CLAUDE.md
git:20260902.4a17ca8 · audit B · 6 stars
cat-mode SKILL.md skill
edbertchan/catstack · corpus/skills/cat-mode/SKILL.md · Edbert's personal working conventions, mined from real session history across his projects. Use when Edbert asks to…
git:20261022.e39bf1c · audit A · 6 stars
principle-assert-invariants-not-last-bug SKILL.md skill
edbertchan/catstack · corpus/skills/principle-assert-invariants-not-last-bug/SKILL.md · Apply after a bug ships past review or a passing grade/check: assert the general class of failure instead of patching…
git:20260915.8e3bf4e · audit A · 6 stars
principle-bind-to-named-inventory SKILL.md skill
edbertchan/catstack · corpus/skills/principle-bind-to-named-inventory/SKILL.md · Apply after an audit, scoring review, or a request for a "real classifier" / "less arbitrary weights" when existing…
git:20260910.26ee82b · audit A · 6 stars
principle-build-the-lever SKILL.md skill
edbertchan/catstack · corpus/skills/principle-build-the-lever/SKILL.md · Apply to any non-trivial work, not just bulk work: edits, migrations, analyses, checks. Build the tool that does it or…
git:20260901.b6a87bb · audit A · 6 stars
principle-encode-lessons-in-structure SKILL.md skill
edbertchan/catstack · corpus/skills/principle-encode-lessons-in-structure/SKILL.md · Apply when you catch yourself writing the same instruction a second time, or notice a recurring correction. Encode the…
git:20260825.c0551cc · audit A · 6 stars
principle-experience-first SKILL.md skill
edbertchan/catstack · corpus/skills/principle-experience-first/SKILL.md · Apply when product, UX, or feature-scope tradeoffs come up. Choose user delight over implementation convenience; ship…
git:20260825.847374d · audit A · 6 stars
principle-explicit-errors SKILL.md skill
edbertchan/catstack · corpus/skills/principle-explicit-errors/SKILL.md · Apply whenever code handles an exception (catch, suppress, translate, retry), decides to skip, drop, truncate, or stop…
git:20260929.5fa51a4 · audit A · 6 stars
principle-fix-root-causes SKILL.md skill
edbertchan/catstack · corpus/skills/principle-fix-root-causes/SKILL.md · Apply when debugging. Trace each symptom to its root cause and fix it there; reproduce first, ask why until you reach…
git:20260825.847374d · audit A · 6 stars
principle-flag-your-own-corrections SKILL.md skill
edbertchan/catstack · corpus/skills/principle-flag-your-own-corrections/SKILL.md · Apply whenever you are about to admit fault in any wording — a fact, number, claim, or check you already stated turns…
git:20260925.a6edf2f · audit A · 6 stars
principle-foundational-thinking SKILL.md skill
edbertchan/catstack · corpus/skills/principle-foundational-thinking/SKILL.md · Apply before writing logic: choosing core types and data structures, sequencing scaffold-vs-feature work, asking what…
git:20260825.847374d · audit A · 6 stars
principle-generalize-from-rejection SKILL.md skill
edbertchan/catstack · corpus/skills/principle-generalize-from-rejection/SKILL.md · Apply when a mechanical gate — a lint, validator, blocked-command guard, or CI check — rejects your output for the same…
git:20260825.847374d · audit A · 6 stars

Every file in edbertchan/catstack

Browse by kind, by grade A, or by owner.