Skip to content

Self-evolving second brain with 35 AI skills, 10 agents, and people CRM. Closed-loop harness: a V-model verification lifecycle where the worker never grades its own homework. Plus paired anti-slop design skills for marketing and product UI. Works with Claude Code, Cursor, Kiro, Gemini CLI, Codex.

docs(claude-md): add engineering discipline protocol (v3.7.1)

Huy Tieudfbf9e9dbaee
diff --git a/.claude-plugin/plugin.json b/.claude-plugin/plugin.json
index 41db627..ac8c4b6 100644
--- a/.claude-plugin/plugin.json
+++ b/.claude-plugin/plugin.json
@@ -1,6 +1,6 @@
{
"name": "cog-second-brain",
- "version": "3.7.0",
+ "version": "3.7.1",
"description": "Self-evolving second brain with 21 AI agent skills, 6 worker agents, and people CRM. Runtime-trust release: memory hygiene sweeps (stale-but-confident), skill post-condition checks (confident-but-unchecked), verbatim citations with adversarial verifier pass, and an autonomous content factory. Iterative skills are loop-engineered (deterministic verifiers, layered termination, in-loop context management). Inspired by Garry Tan's gstack specialist sessions and gbrain knowledge patterns. Works with Claude Code, Cursor, Kiro, Gemini CLI, Codex, and any markdown-reading agent.",
"author": {
"name": "Huy Tieu",
diff --git a/.cursor-plugin/plugin.json b/.cursor-plugin/plugin.json
index 860cee0..412f959 100644
--- a/.cursor-plugin/plugin.json
+++ b/.cursor-plugin/plugin.json
@@ -1,7 +1,7 @@
{
"name": "cog-second-brain",
"displayName": "COG: Agentic Second Brain",
- "version": "3.7.0",
+ "version": "3.7.1",
"description": "Self-evolving second brain powered by AI agents, Obsidian markdown, and Git. 21 skills, 6 worker agents, people CRM with tiered enrichment. Runtime-trust release: memory hygiene sweeps, skill post-condition checks, verbatim citations with verifier pass, autonomous content factory. Iterative skills are loop-engineered with deterministic verifiers and layered termination. Inspired by Garry Tan's gstack and gbrain patterns. Works with Claude Code, Cursor, Kiro, Gemini CLI, Codex, and any markdown-reading agent.",
"author": {
"name": "Huy Tieu",
diff --git a/CHANGELOG.md b/CHANGELOG.md
index 4206a96..f360649 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -2,6 +2,21 @@
All notable changes to COG (Cognition + Obsidian + Git) will be documented in this file.
+## [3.7.1] - 2026-07-10
+
+### Added
+
+#### Engineering Discipline protocol in CLAUDE.md
+Operating rules for agents doing engineering work through COG:
+
+- **Code comments** — no decorative comment separator blocks (`// ====`, `/* ==== Section ==== */`, full-line dividers); plain comments and blank lines separate sections.
+- **Git** — never `git reset --hard` or `git commit --amend` unless explicitly asked; new commits only, recover via `git reflog`. Commitlint / Conventional Commits standards. Non-interactive flags (`GIT_EDITOR=true`, `--no-edit`) for commands that would prompt.
+- **Pull requests** — check for and follow the repo's PR template; review replies in-thread via `gh api .../pulls/comments/{id}/replies` (never a new parent comment), resolve threads via GraphQL `resolveReviewThread`, no pleasantries — state what changed, which commit, and why.
+- **Interaction** — read every user-provided file before responding; answer explanatory questions verbally before invoking tools, and wait for an explicit request before investigating or changing code.
+
+### Changed
+- Version bumped to **3.7.1** across `COG-VERSION` and all plugin manifests.
+
## [3.7.0] - 2026-07-10
### Runtime Trust
diff --git a/CLAUDE.md b/CLAUDE.md
index cda4356..41efd42 100644
--- a/CLAUDE.md
+++ b/CLAUDE.md
@@ -87,6 +87,26 @@ This is opt-in per skill — not a hard rule for every output. Apply it where wr
---
+## Engineering Discipline — ALWAYS APPLY
+
+### Code Comments
+- Never use decorative comment separator blocks of any kind — `// ====`, `// ----`, `// ---- Section Name ----`, `// -----------` full-line dividers, `/* ==== Section Name ==== */`, etc. Use plain single-line comments and blank lines to separate sections instead.
+
+### Git
+- Never `git reset --hard` or `git commit --amend` unless the user explicitly asks. Always create new commits and push normally. If changes get wiped, recover from `git reflog` — do not destroy history.
+- Always commit with commitlint standards (Conventional Commits: `type(scope): subject`).
+- When a command requires interactive input (e.g. `git rebase --continue`, editor prompts), supply the non-interactive flag or set `GIT_EDITOR=true` / `EDITOR=true` / `--no-edit` as appropriate.
+
+### Pull Requests
+- Before opening a PR, check the repository for a PR template (e.g. `.github/PULL_REQUEST_TEMPLATE.md` or similar) and always follow it when composing the PR description.
+- PR review replies must be in-thread via `gh api repos/{owner}/{repo}/pulls/comments/{id}/replies`, never a new parent comment, then resolve the thread via the GraphQL `resolveReviewThread` mutation. No pleasantries ("great catch", etc.) — state what changed, which commit, and why.
+
+### Interaction
+- Read every file the user provides (images, screenshots, code, text) with the read tool before responding — never assume its contents.
+- Answer the user before acting. Explanatory questions should be answered verbally without first invoking tools or editing code. Wait for the user to explicitly request investigation, a fix, or changes.
+
+---
+
## Skill Post-Condition Rule — ALWAYS APPLY
Every skill run that **mutates external state** (publishes a page, deploys, posts to Slack/socials, transitions a ticket, pushes commits, fires a webhook) must end with an explicit **post-condition check**: fetch back or observe the mutated artifact and confirm it matches intent before reporting success.
diff --git a/COG-VERSION b/COG-VERSION
index 7c69a55..a76ccff 100644
--- a/COG-VERSION
+++ b/COG-VERSION
@@ -1 +1 @@
-3.7.0
+3.7.1
diff --git a/marketplace-entry.json b/marketplace-entry.json
index 7653a39..18a2b46 100644
--- a/marketplace-entry.json
+++ b/marketplace-entry.json
@@ -7,7 +7,7 @@
"url": "https://github.com/huytieu"
},
"description": "Self-evolving second brain with 21 AI skills, 6 worker agents, and people CRM. Runtime-trust release: memory hygiene sweeps, skill post-condition checks, verbatim citations with adversarial verifier pass, autonomous content factory. Iterative skills are loop-engineered with deterministic verifiers and layered termination. Inspired by Garry Tan's gstack and gbrain. Works with Claude Code, Cursor, Kiro, Gemini CLI, Codex.",
- "version": "3.7.0",
+ "version": "3.7.1",
"keywords": [
"second-brain",
"garry-tan",