v0.9.4

Human-first README rewrite — the npm package page now renders a proper README explaining what cue is, who it's for, and how to get started. All links converted to absolute URLs so they work on npmjs.com; fixed stale .cue-profile → .cue.profile reference. See #59.

v0.9.3

What's Changed

• fix(ci): bump skills submodule gitlink to resolve E3 SKILL_NOT_FOUND by @NagyVikt in #39
• ci(profiles): watch resources/skills + resources/mcps gitlinks by @NagyVikt in #40
• feat: free accounts + API tokens for cuecards.cc (BetterAuth + Neon) by @NagyVikt in #41
• feat(security): wire trivy-scan supply-chain gate into core by @NagyVikt in #43
• feat(summon): cue summon — bind a profile into the live session, no restart by @NagyVikt in #42
• Reduce default cue context overhead by @NagyVikt in #45
• Prune default core MCPs by @NagyVikt in #46
• Feat/cue summon by @NagyVikt in #47
• fix: trim meta-descriptions to <=155 chars (SEO best practice) by @NagyVikt in #48
• perf(profiles): cut per-session CLAUDE.md tokens ~25% across all profiles by @NagyVikt in #49
• perf(core): token-effective profiles — model routing by task hardness by @NagyVikt in #50
• test: fix flaky dashboard-merge test + assert codex persona parity by @NagyVikt in #52
• fix(ci): remove stray hirelens-showcase gitlink breaking submodule checkout by @NagyVikt in #51
• fix(hooks): parse multi-line -m commit messages in commit-message-guard by @NagyVikt in #54
• fix(runtime): account identity follows CLAUDE_CONFIG_DIR on shared profile runtimes by @NagyVikt in #55
• fix(runtime): rescue login-fresh credentials before the identity guard wipes them by @NagyVikt in #56
• Fix/ci hirelens gitlink by @NagyVikt in #57
• chore(release): bump version to 0.9.3 by @NagyVikt in #58

Full Changelog: v0.9.2...v0.9.3

v0.9.2 — live review visibility + self-learner

Two opt-in capabilities, plus docs for the review gate.

🔭 Live code-review visibility

Watch an independent review move file-by-file in real time instead of an opaque "Precipitating…" spinner.

• cue-review-watch — live renderer; run it in a second pane to follow a review as it happens.
• cue-review-progress — append-only progress events to ~/.config/cue/review-progress/<id>.jsonl (the shared schema every reviewer writes).
• /code-review emits per-file / per-dimension / per-finding events.
• The auto-review Stop-hook gate records its review to the same log — verdict is parsed with the progress side-channel filtered out, so verdict integrity never depends on a pipe and progress text can't be misread as a verdict.
• Docs: docs/review-visibility.md.

🌱 Self-learner (experimental · opt-in · default-OFF)

Profiles capture where their skills fell short during a task and feed gated improvements back over time.

• profile-self-improve.sh — friction-signal capture plus an optional live critic agent. Recursion-guarded, never blocks Stop, runs the critic at most once per session, fully fail-open.
• New skill_gap analytics event (inert to existing readers). Piloted on the skill-writer profile.
• Enable with touch ~/.config/cue/.auto-improve-enabled. Docs: docs/self-learner.md.

📖 Docs

• README — new "what you'll see during a run — the reviewer" section: why a red "Stop hook error" means the review gate is working, and how to watch a review live.

---

Tests: review-progress 5/5 · profile-self-improve 8/8 · auto-review 6/6 (no regression) · typecheck + biome lint clean.

Full changelog: CHANGELOG.md · merged in #38.

v0.9.1 — cue studio dashboard, lean install, profile + CLI fixes

What's new in v0.9.1

Point release over v0.9.0. The headline arrivals: cue studio matures into a real dashboard/IDE, a new lean install path, and a batch of profile + CLI fixes (PRs #6–#33).

🖥️ cue studio dashboard

• Environment Variables view — per-folder .env, masked secrets, transient reveal, with security hardening from review (#20, #23).
• Hooks / permissions rail + live SSE activity chart + inline hook-source viewer (#27, #30).
• profile-linter, repos catalog, and profile tuning (#32).
• cue dash CLI surface; Active Sessions card polish (#30, #13).

🪶 Lean install path

• New setup/lean-cue.md — cross-OS (Linux / macOS / Windows-via-WSL2) minimal stack: cue CLI + core profile + caveman + RTK, nothing else (#29).
• Wired into README + AGENTS.md discovery; RTK x86_64/ARM + Homebrew fallback notes across all setup docs (#31).

🧩 Profiles & skills

• vercel profile: vercel-labs/agent-skills + agent-browser (#28).
• browser profile: playwright (#18).
• cue-developer skill via submodule bump (#33).

🔧 CLI & validation

• Lazy-require bun:sqlite so the node-bundled CLI boots under plain node (#22).
• cue validate: private/local-only MCP refs demoted from E3 → W9 (#25).
• Gated autoship — gitguardex create→review→merge (#16).
• Grouped combine picker (categories, headers, danger tag) (#15).
• trigger-gaps word-boundary + hang fixes (#12, #14); analytics day-window off-by-one (#18).

Install / upgrade

```
npm install -g cue-ai@0.9.1
```

Full changelog: v0.9.0...v0.9.1

v0.9.0 — lean profiles, cuecards rename, core hardening

What's new in v0.9.0

The lean release: trim the always-on profiles, formalize the cuecards rename, and harden core with verification hooks.

🪶 Leaner always-on profiles

The big change — core and gstack were carrying a long tail of rarely-used skills on every session. Cut hard, kept the high-value roles, and made everything else load on demand.

• core: 27 → 15 skills. Dropped 0-telemetry / non-universal skills: meta/just, meta/kiro-powers, meta/mcp-finder, plan/office-hours, plan/plan-ceo-review, plan/plan-eng-review, review/code-review (dup of the command + code-review-deep), gstack/context-save, gstack/context-restore, security/cso, tools/learn-from-repo, tools/prior-art. These now soft-load on demand via meta/smart-loader when the topic matches.
• gstack: 51 → 14 skills. Kept the high-value roles — ship/deploy (ship, land-and-deploy, setup-deploy, canary, document-release), design + real-browser QA (design-review, design-shotgun, devex-review, qa, qa-only), plan/spec (spec, plan-design-review, plan-devex-review), and health. Dropped the long tail: iOS×5, benchmark×2, browse, codex, retro, autoplan, office-hours, plan-ceo/eng-review, scrape, skillify, setup-gbrain, sync-gbrain, make-pdf, learn, design-html, landing-report, and more.
• full profile now homes every cut skill via the */* glob, so cue doctor stays orphan-free and any new skill is auto-covered.

✨ Core hardening

• liedetector /verify command wired into core, with tag-audit and brain-next-suggestion Stop hooks for verification-grounded output.
• smart-loader UserPromptSubmit hook surfaces relevant off-profile skills proactively.
• review/code-review promoted into core (was duplicated across 10 profiles).
• SkillNotFound now suggests <category>/<name> for bare slugs.

📮 Profile updates

• postizz: added the 4 first-party Higgsfield image/video skills + the Higgsfield MCP for brand asset generation.
• marketing: declares the trendradar MCP + content skills.
• backend / postgres: wire their MCP servers.

⚡ Performance & CI

• Launch speed: the launch-time description lint no longer shells npx skills add into a throwaway temp dir on every rebuild — npx-heavy profile launches (e.g. postizz) drop from ~33s to ~1.4s. cue validate still does the full online check.
• Green CI: repaired the test suite (stale ecc fixtures → gstack, full profile globbing, glob-aware cue doctor) and rewrote the e2e doctor-drift + npx scenarios for the current architecture.

📦 Rename to cuecards

Package, README, and GitHub Pages theme renamed to cuecards.

Install / upgrade

npm install -g cue-ai@0.9.0

v0.8.0 — gstack skills, claude-ads integration, profile logos

What's new in v0.8.0

🎯 Full claude-ads integration (google-ads profile)

• 23 sub-skills from AgriciDaniel/claude-ads v1.7.1
• 250+ audit checks across 8 ad platforms (Google, Meta, YouTube, LinkedIn, TikTok, Microsoft, Apple, Amazon)
• Live GAQL API queries via official Google Ads MCP + multi-platform audit mode

🛠️ 36 gstack skills ported

• core: context-save, context-restore
• ecc: autoplan, benchmark-models, codex, health, investigate, office-hours, plan-ceo/eng/devex-review, retro, review, ship, land-and-deploy
• backend: canary, health, investigate, landing-report, setup-deploy, ship, review, land-and-deploy, retro, cso
• frontend: benchmark, canary, design-consultation, design-review, plan-design-review, scrape, skillify, setup-browser-cookies
• docs-writer: document-generate, document-release, make-pdf
• cybersecurity: cso (Chief Security Officer mode)
• research: scrape, skillify, hackernews-frontpage

🖼️ Profile logos (Kitty terminal)

• google-ads: Google Ads icon
• instagram: official IG gradient logo
• postizz: official Postiz logo (from postiz.com)
• medusa-dev: Medusa JS hexagon logo

🔧 Improvements

• Picker outro now shows profile icon (profile: 🐢 core)
• cli-recipes.json: added scc and playwright recipes
• Lint-fixed all 36 gstack skills (71 auto-fixes)
• README profile catalog regenerated with updated skill counts

📦 Install / upgrade

npm install -g cue-ai@0.8.0