Release v0.5.4

.agents/plugins/marketplace.json

@@ -6,7 +6,7 @@
   "plugins": [
     {
       "name": "slow-powers",
-      "version": "0.5.3",
+      "version": "0.5.4",
       "source": {
         "source": "url",
         "url": "./"

.claude-plugin/marketplace.json

@@ -9,7 +9,7 @@
     {
       "name": "slow-powers",
       "description": "Agent skills to upgrade plan mode and debugging, enforce best practices, and enhance native features, not replace them.",
-      "version": "0.5.3",
+      "version": "0.5.4",
       "source": "./",
       "author": {
         "name": "Max Haarhaus",

.claude-plugin/plugin.json

@@ -1,7 +1,7 @@
 {
   "name": "slow-powers",
   "description": "Agent skills to upgrade plan mode and debugging, enforce best practices, and enhance native features, not replace them.",
-  "version": "0.5.3",
+  "version": "0.5.4",
   "author": {
     "name": "Max Haarhaus",
     "email": "samiamorwas@gmail.com"

.codex-plugin/plugin.json

@@ -1,6 +1,6 @@
 {
   "name": "slow-powers",
-  "version": "0.5.3",
+  "version": "0.5.4",
   "description": "Agent skills to upgrade plan mode and debugging, enforce best practices, and enhance native features, not replace them.",
   "author": {
     "name": "Max Haarhaus",
@@ -37,8 +37,8 @@
     "privacyPolicyURL": "https://docs.github.com/en/site-policy/privacy-policies/github-general-privacy-statement",
     "termsOfServiceURL": "https://docs.github.com/en/site-policy/github-terms/github-terms-of-service",
     "brandColor": "#FFF8F5",
-    "composerIcon": "./assets/slow-powers-small.svg",
-    "logo": "./assets/app-icon.png",
+    "composerIcon": "./assets/mark.png",
+    "logo": "./assets/logo.png",
     "screenshots": []
   }
 }

.github/badges/evals.json

@@ -0,0 +1,6 @@
+{
+  "schemaVersion": 1,
+  "label": "evals",
+  "message": "measuring",
+  "color": "lightgrey"
+}

.gitignore

@@ -1,12 +1,7 @@
 .worktrees/
-.private-journal/
 # Ignore local Claude Code state, but track the shared, committed settings.
 .claude/*
 !.claude/settings.json
 .DS_Store
 node_modules/
-inspo
-triage/
-# Ephemeral eval run output. Curated baselines are promoted into
-# skills/<skill>/evals/baseline/ (tracked) via `bun run evals:promote-baseline`.
-skills-workspace/
+.eval-magic

AGENTS.md

@@ -11,8 +11,6 @@ This repo ships Slow-powers across three harnesses:
 - `.codex-plugin/` — OpenAI Codex plugin
 - `opencode/` — OpenCode plugin (`@slowdini/slow-powers-opencode`)
 
-See the [feature support](README.md#feature-support) tables in the README for current tier per harness.
-
 ## Editing the right files
 
 Two file-confusion traps are common in this repo. Avoid both:
@@ -33,7 +31,6 @@ All skills MUST use cross-harness vocabulary, as described in `slow-powers/writi
 
 ## Pull Request Requirements
 
-- One problem per PR. Bundled unrelated changes will be split or sent back.
 - Read existing skills before proposing changes to skill content. Skill
   prose has been tuned over many iterations upstream and downstream; changes
   to behavior-shaping content (Red Flags tables, rationalization lists,

CONTRIBUTING.md

@@ -0,0 +1,32 @@
+# Contributing to Slow-powers
+
+Thanks for your interest in Slow-powers! The best way to contribute right now is
+to **open an issue** — bug reports, workflow feedback, and ideas are all genuinely
+helpful and shape where the project goes.
+
+## Open an issue
+
+Head to [**New issue**](https://github.com/slowdini/slow-powers/issues/new/choose)
+and pick the template that fits:
+
+- **Mechanical Bug Report** — crashes, install failures, plugin load errors, or
+  other things that are plain broken.
+- **Workflow Bug Report** — the agent skips a skill, uses the wrong one, or
+  misreads instructions. Include the skill(s) involved and a transcript excerpt.
+- **Other** — questions, documentation issues, or general feedback.
+
+Filling in the template (version, harness, model, and a clear repro) is the single
+biggest thing you can do to help us act on a report quickly.
+
+## Pull requests
+
+We are **not accepting unsolicited community pull requests** at this time.
+Slow-powers' skill prose is tuned through repeated evaluation, and changes need
+context and eval evidence that a drive-by PR can't carry. If you've spotted a
+problem or have a change in mind, please **open an issue** describing it — that's
+the path that gets things fixed.
+
+## Code of Conduct
+
+This project is governed by our [Code of Conduct](./CODE_OF_CONDUCT.md). By
+participating, you're expected to uphold it.

README.md

@@ -1,7 +1,39 @@
+<p align="center">
+  <img src="assets/logo.png" alt="Slow-powers" width="360" />
+</p>
+
+<p align="center">
+  <a href="https://github.com/slowdini/slow-powers/actions/workflows/ci.yml"><img
+    src="https://img.shields.io/github/actions/workflow/status/slowdini/slow-powers/ci.yml?branch=main"
+    alt="CI status" /></a>
+  <a href="https://github.com/slowdini/slow-powers/releases/latest"><img
+    src="https://img.shields.io/github/v/release/slowdini/slow-powers"
+    alt="Latest release" /></a>
+  <a href="https://www.npmjs.com/package/@slowdini/slow-powers-opencode"><img
+    src="https://img.shields.io/npm/v/@slowdini/slow-powers-opencode"
+    alt="OpenCode plugin on npm" /></a>
+  <a href="./LICENSE"><img
+    src="https://img.shields.io/github/license/slowdini/slow-powers"
+    alt="License: MIT" /></a>
+  <a href="#why-trust-these-skills"><img
+    src="https://img.shields.io/endpoint?url=https%3A%2F%2Fraw.githubusercontent.com%2Fslowdini%2Fslow-powers%2Fdev%2F.github%2Fbadges%2Fevals.json"
+    alt="evals" /></a>
+</p>
+
 # Slow-powers
 
 Slow-powers is an agent skill set for professional software development. It enhances plan mode and debugging work, enforces best practices, and works _with_ the features of modern agents, instead of replacing them. It's the plugin for people who don't install plugins.
 
+<p align="center">
+  <img src="assets/demo/tdd-demo.gif"
+    alt="The test-driven-development skill in a coding agent: write a failing test, watch it fail for the right reason, write minimal code — and the test catches a race before it ships."
+    width="760" />
+</p>
+
+<p align="center">
+  <sub>The <code>test-driven-development</code> skill catching a race before it ships. Terminal themed with <a href="https://github.com/samiamorwas/synthpunk">Synthpunk Neon Dark</a>.</sub>
+</p>
+
 ## About this fork
 
 Slow-powers is a fork of [obra/superpowers](https://github.com/obra/superpowers). Much of the skill content is sourced from upstream, with rewrites focusing on clarity, token efficiency, and providing a generally lighter touch.
@@ -76,6 +108,19 @@ Slow-powers provides a set of highly focused skills that ensure your agent opera
 6. **`writing-skills`** — Helps write and edit skills, following the same best practices that guide slow-powers itself.
 7. **`evaluating-skills`** — Teaches the agent how to run skill evals, so the value of skills and prose changes can be objectively assessed.
 
+## Why trust these skills?
+
+Most skill packs ship on vibes. Every slow-powers skill ships with a documented eval — or it doesn't ship (see [Philosophy](#philosophy)). Each skill is measured against an agent with **no skill**, so the number means "this skill made the agent better," not just "we changed something."
+
+| Skill | Improvement vs no skill | n | Model | Last measured |
+|---|---|---|---|---|
+| `hardening-plans` | TBD | TBD | TBD | TBD |
+| `investigating-bugs` | TBD | TBD | TBD | TBD |
+| `test-driven-development` | TBD | TBD | TBD | TBD |
+| `verifying-development-work` | TBD | TBD | TBD | TBD |
+
+**Improvement vs no skill** is the gain in eval pass-rate (percentage points) when the same [eval-magic](https://github.com/slowdini/eval-magic) suite runs with the skill versus without it.
+
 ## Intended Workflows
 
 The skills declare prerequisite / next-step gates so the agent follows an intended skill sequence. These gates **suggest** what comes before and after a skill once it is invoked; they do **not** restrict when any skill can be invoked.

assets/demo/tdd-playback.sh

@@ -0,0 +1,73 @@
+#!/usr/bin/env bash
+# Re-enactment of a REAL Claude Code test-driven-development session (RED -> GREEN).
+#
+# Deterministic: pacing lives here (the "actor"), VHS is the "camera".
+# Colors use base-16 ANSI (30-37 / 90-97 + bold) so the terminal theme drives them
+# (here, Synthpunk Neon Dark). Glyphs limited to VHS's default font: "└─ ✓ ✗ ●".
+set -u
+
+RST=$'\e[0m'; B=$'\e[1m'
+FG=$'\e[97m'     # brightWhite  -> readable text
+DIM=$'\e[90m'    # brightBlack  -> muted chrome
+RED=$'\e[31m'    # fail
+GREEN=$'\e[32m'  # pass
+YEL=$'\e[33m'    # agent bullet
+CYAN=$'\e[36m'   # skill name
+BLUE=$'\e[34m'   # prompt caret
+
+human()  { printf '%s>%s %s%s%s\n' "$BLUE" "$RST" "$B$FG" "$1" "$RST"; }
+bullet() { printf '\n%s●%s %s\n' "$YEL" "$RST" "$1"; }
+tool()   { printf '\n%s●%s %s%s%s%s(%s)%s\n' "$YEL" "$RST" "$B$FG" "$1" "$RST" "$DIM" "$2" "$RST"; }
+res()    { printf '%s  └─%s %s\n' "$DIM" "$RST" "$1"; }
+cont()   { printf '%s     %s%s\n' "$DIM" "$1" "$RST"; }
+pause()  { sleep "$1"; }
+
+clear
+pause 0.4
+
+# the human's request (generic: a data-fetching change with a subtle ordering bug)
+human "for your own account, read the api key from the session instead of refetching it"
+pause 0.9
+
+bullet "Using ${B}${CYAN}test-driven-development${RST}."
+pause 0.8
+
+# ---------- RED ----------
+printf '\n%s  %sRED%s — write the failing test first\n' "$B" "$RED" "$RST"
+pause 0.5
+tool "Write" "overview.test.ts"
+res  "${FG}it(\"reads the key from the session for your own account\")${RST}"
+pause 0.9
+tool "Bash" "bun test"
+pause 0.7
+res  "${RED}✗ reads the key from the session for your own account${RST}"
+cont "${DIM}expected the session key, but fetchAccount() was called${RST}"
+res  "${RED}0 pass, 1 fail${RST}"
+pause 0.6
+printf '  %s✓%s fails for the right reason — not wired up yet\n' "$GREEN" "$RST"
+pause 1.1
+
+# ---------- GREEN (first attempt reveals a race the test catches) ----------
+printf '\n%s  %sGREEN%s — minimal code to pass\n' "$B" "$GREEN" "$RST"
+pause 0.5
+tool "Edit" "overview.ts"
+res  "${FG}use the session key when it's your own account${RST}"
+pause 0.9
+tool "Bash" "bun test"
+pause 0.7
+res  "${RED}✗ fetchAccount() still fired once${RST}"
+cont "${DIM}race: it ran before the session finished loading${RST}"
+pause 0.8
+bullet "The test caught a race the first cut missed — fixing before I move on."
+pause 1.1
+tool "Edit" "overview.ts"
+res  "${FG}gate the fetch on the session being loaded${RST}"
+pause 0.9
+tool "Bash" "bun test"
+pause 0.7
+res  "${GREEN}✓ reads the key from the session for your own account${RST}  ${DIM}[3ms]${RST}"
+res  "${GREEN}12 pass, 0 fail${RST}"
+pause 0.7
+
+bullet "${B}${GREEN}Green.${RST} The test caught a race before it shipped."
+pause 2.0

assets/demo/tdd.tape

@@ -0,0 +1,27 @@
+# VHS tape for the slow-powers CLI demo (see README hero GIF).
+# Regenerate with:  bun run demo   (which runs:  vhs assets/demo/tdd.tape)
+# Paths are relative to the repo root, where vhs is invoked.
+#
+# The pacing/content lives in the actor script (tdd-playback.sh); this tape is
+# just the "camera": canvas, font, and terminal theme.
+
+Output assets/demo/tdd-demo.gif
+Set Shell bash
+
+# Terminal theme: Synthpunk Neon Dark — github.com/samiamorwas/synthpunk
+# (VHS takes a named theme or inline JSON, not a file path, so it is inlined here.)
+Set Theme { "name": "Synthpunk Neon Dark", "background": "#0F0620", "foreground": "#E8E0F0", "cursor": "#D040FF", "selection": "#123A52", "black": "#060210", "red": "#FF3A50", "green": "#40FF80", "yellow": "#FFE040", "blue": "#A060FF", "magenta": "#FF2A8A", "cyan": "#00E8F0", "white": "#B8A8D0", "brightBlack": "#8878A0", "brightRed": "#FF6173", "brightGreen": "#66FF99", "brightYellow": "#FFE666", "brightBlue": "#B380FF", "brightMagenta": "#FF55A1", "brightCyan": "#27F8FF", "brightWhite": "#E8E0F0" }
+
+Set FontSize 19
+Set Width 1080
+Set Height 900
+Set Padding 30
+Set Framerate 24
+Set TypingSpeed 0ms
+
+Hide
+Type "bash assets/demo/tdd-playback.sh"
+Enter
+Sleep 300ms
+Show
+Sleep 15s

package.json

@@ -1,6 +1,6 @@
 {
   "name": "@slowdini/slow-powers-opencode",
-  "version": "0.5.3",
+  "version": "0.5.4",
   "description": "Slow-powers — structured development workflows for coding agents (TDD, debugging, verification, git hygiene)",
   "type": "module",
   "main": "./opencode/plugins/slow-powers.js",
@@ -49,6 +49,7 @@
     "evals:grade": "eval-magic grade --skill-dir ./skills",
     "evals:aggregate": "eval-magic aggregate --skill-dir ./skills",
     "evals:promote-baseline": "eval-magic promote-baseline --skill-dir ./skills",
+    "demo": "vhs assets/demo/tdd.tape",
     "check": "biome check --write .",
     "check:ci": "biome check --error-on-warnings .",
     "typecheck": "tsc --noEmit",

skills/evaluating-skills/SKILL.md

@@ -33,7 +33,7 @@ A **deterministic** change doesn't move that needle. Removing a one-line "announ
 
 **Either way, announce the decision and why** — "deterministic instruction removal, no eval" or "this changes pressured compliance, I'll run an eval." A visible decision is one the user can override; a silent one is a rationalization waiting to happen. **The door stays open:** if the user wants an eval anyway, run a worthwhile one — design real cases, don't phone it in to confirm a foregone conclusion.
 
-**Skill type is a fast read, not a verdict.** Reference and manually-invoked procedural changes *often* land deterministic; discipline, technique, and pattern changes *often* carry contingency (`pressure-scenarios.md` draws the same line under "When to use" / "Don't use them for"). Use type to orient your first guess — never as the answer. The decision is per *change*, not per type: a deterministic typo fix in a discipline-enforcing skill still skips, and restructuring a reference doc because the agent kept missing a section is contingent and earns an eval.
+**Skill type is a fast read, not a verdict.** Reference and manually-invoked procedural changes *often* land deterministic; discipline, technique, and pattern changes *often* carry contingency ([`pressure-scenarios.md`](references/pressure-scenarios.md) draws the same line under "When to use" / "Don't use them for"). Use type to orient your first guess — never as the answer. The decision is per *change*, not per type: a deterministic typo fix in a discipline-enforcing skill still skips, and restructuring a reference doc because the agent kept missing a section is contingent and earns an eval.
 
 ## The Iron Law
 
@@ -103,15 +103,15 @@ Tips for writing good prompts:
 - **Vary phrasing.** Mix casual ("hey can you check this") with precise ("Run `bun test`, quote the output").
 - **Cover edge cases.** Include at least one boundary condition, malformed input, or ambiguous instruction.
 - **Use realistic context.** Real users reference file paths, function names, personal context. "Process this data" is too vague to test anything useful.
-- **For discipline-enforcing skills**, see `pressure-scenarios.md` for the pressure-scenario taxonomy (time pressure, sunk cost, authority, exhaustion, etc.).
+- **For discipline-enforcing skills**, see the [pressure-scenario taxonomy](references/pressure-scenarios.md) (time pressure, sunk cost, authority, exhaustion, etc.).
 
 **Don't write assertions yet.** You don't know what "good" looks like until you see what the first run produces.
 
 ### Testing by skill type
 
 What "stresses the skill" depends on what kind of skill it is. The four types from `slow-powers:writing-skills` each need a different style of prompt:
 
-- **Discipline-enforcing skills** (TDD, verifying-development-work). Test with pressure — academic prompts ("explain how TDD works") will pass without measuring anything useful. Combine multiple pressures (time + sunk cost + authority + exhaustion) and force a choice. See `pressure-scenarios.md` for the taxonomy. The wild failure for these skills is almost always *mid-session* — the agent is already committed to a skill-free approach when the trigger arrives — so a cold prompt under-measures them; pair each cold case with a **seeded** one (see *Seeding conversation context* below). Success = the rule holds under maximum pressure.
+- **Discipline-enforcing skills** (TDD, verifying-development-work). Test with pressure — academic prompts ("explain how TDD works") will pass without measuring anything useful. Combine multiple pressures (time + sunk cost + authority + exhaustion) and force a choice. See the [pressure-scenario taxonomy](references/pressure-scenarios.md). The wild failure for these skills is almost always *mid-session* — the agent is already committed to a skill-free approach when the trigger arrives — so a cold prompt under-measures them; pair each cold case with a **seeded** one (see *Seeding conversation context* below). Success = the rule holds under maximum pressure.
 - **Technique skills** (condition-based-waiting, root-cause-tracing). Test application: hand the agent a new scenario where the technique applies and check it gets used correctly. Include at least one edge-case variation. Success = the technique transfers to a situation the skill didn't explicitly describe.
 - **Pattern skills** (flatten-with-flags, information-hiding). Test recognition: include prompts where the pattern applies and prompts where it doesn't. Success = the agent applies the pattern when warranted and refrains when it isn't.
 - **Reference skills** (API docs, syntax guides). Test retrieval: ask questions whose answers are in the reference, including a few that hit gaps you suspect. Success = the agent finds the right section and uses it correctly.
@@ -199,6 +199,6 @@ The mechanics of executing a run live in **[eval-magic](https://github.com/slowd
 ## See also
 
 - `slow-powers:writing-skills` — drafting a skill (Phase 1)
-- `pressure-scenarios.md` — pressure-scenario taxonomy for authoring prompts that stress discipline-enforcing skills
+- [Pressure scenarios](references/pressure-scenarios.md) — taxonomy for authoring prompts that stress discipline-enforcing skills
 - eval-magic (the `eval-magic` tool) — runs the evals this skill teaches you to author
 - agentskills.io/skill-creation/evaluating-skills — the methodology this skill is derived from