docs/listing/release-notes.md

# Release Notes

## 0.6.0

One skill and one command on every platform: `all-the-medicine`, now always on through a Claude SessionStart hook, a Codex `AGENTS.md` block and a Cursor rule. The individual controls are references inside it, so menus show one entry instead of 31. Adds a one-line installer for Claude Code, Codex and Cursor.

## 0.5.0

Adds `direct-communication` for terse, assertive answers, plain language, essential evidence, and one specific non-looping blocker question. It replaces the installed `communicate-briefly` control, is callable in Claude and Codex, and is automatically compiled into `all-the-medicine`. Codex starter prompts now satisfy the maximum count and 128-character limit.

## 0.4.0

Adds two explicit-invocation "superpower" skills. `never-stop` provides maximum autonomous execution and `all-the-medicine` dynamically orchestrates every registered public skill with deterministic conflict handling and self-recursion protection.

## 0.3.0

Adds a semantic anti-bypass layer and complementary underthinking/reasoning-balance controls. The release detects false progress, false completion, false blockers, hidden recursion, strategy laundering, scope laundering, memory poisoning, context starvation, shallow debugging, and premature parking. It adds 19 public skills, 15 rules, seven focused guides, compact JSON/TOON/SJON policy state, a 24-loophole catalog, ten underthinking regressions, controlled executive overrides, deterministic rule conflict resolution, and an evidence-based final self-check. Existing anti-overthinking behavior remains intact.

## 0.2.0

Adds direct intervention skills, substantive ADHD/OCD behavioral-analogy guidance, overthinking and recursive-task controls, hallucination/evidence recovery, a complete failure-mode catalog, deterministic state assessment, and exact prompt traceability.

## 0.1.0

Initial cross-platform release with one shared installer, complete master prompt, 41 rules, 28 skills, adapters, context, memory, machine state, traceability, scenarios, and validation.