v12: Open-Core Reset — make every claim true - #32
Open
frankxai wants to merge 7 commits into
Open
Conversation
- Add MIT LICENSE file (was claimed in 4+ places but never shipped; repo was legally all-rights-reserved) - Fix package.json author to Frank Riemer (was invented 'Frank Guzman'), homepage -> frankx.ai/acos, v12.0.0 - Fix acos-health-check.yml empty schedule: key (every push showed red X startup_failure); add manifest validation, leak gate, stats gate - Remove Anthropic-proprietary skill mirrors (.claude/skills/anthropic/ + docx/pptx/xlsx/pdf — '© 2025 Anthropic, PBC. All rights reserved' inside an MIT repo) - Purge 21 claude-mem session-telemetry CLAUDE.md files - Remove Oracle-internal OneDrive path from frankx-content-creation agent - Remove €1000-tier pricing/defensibility strategy lines from public docs/ops - Remove .swarm/ session DB; gitignore .swarm/ and workstation-private registry - Clean settings.json: drop third-party commit attribution (ruv@ruv.net), claude-flow allowlist + inert daemon config - Fix opencode.json dead absolute Windows paths -> relative - Archive BRANCH_AUDIT.md + ecosystem-report.md ops logs Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
- Remove 21 vendored claude-flow agent subdirs (79 files: byzantine/raft/CRDT coordinators, hive-mind, sparc, trading-predictor etc.) — upstream project content, not ACOS
- Remove vendored claude-flow command trees (sparc/, github/, analysis/, automation/, monitoring/, optimization/, hooks/, claude-flow-agents/) + gsd/ foreign system commands and hooks
- Remove vendored gstack skill tree (229 files) per repo's own doctrine ('Do not vendor GStack into ACOS core'); gstack stays an optional external install
- Remove .claude/slash-commands/ (CRLF-broken scripts pointing at dead WSL paths)
- Remove 0-byte nextjs-react-expert skill (active auto-activation target loading nothing)
- Move FrankX-personal surfaces to instances/frankx/: 11 frankx-ai-* commands, frankx-content-creation + weekly-recap agents, 6 personal skills (frankx-brand, frankx-daily-execution, gym/health/spartan/greek)
- Replace all dead /mnt/c/Users/Frank + C:/Users/Frank absolute paths with ~ across 39 files
- Fix agent name collision: two agents both named 'Music Producer' -> music-producer / frequency-music-producer
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
- plugin.json v12: valid schema only (drop invented compose/stats/intelligence fields), declare real component paths (.claude/commands|agents|skills, hooks/hooks.json, .mcp.json)
- Add .claude-plugin/marketplace.json ('frankx' marketplace) so 'claude plugin marketplace add frankxai/agentic-creator-os' works
- New hooks/hooks.json in Claude Code format wiring 3 telemetry-only Node hooks (UserPromptSubmit skill activation, PostToolUse tracking, Stop finalize) via CLAUDE_PLUGIN_ROOT — doctrine-compliant: fast, cross-platform, no quality gates
- FIX skill auto-activation (flagship feature, silently dead): rules path bug + schema mismatch — hook expected promptTriggers schema, rules ship activation_rules[]; added normalizer + robust path resolution. Verified firing end-to-end
- Resolve hooks/ naming collision: marketing hook-writing KB -> content-hooks/, mcp-health-check diagnostic -> scripts/
- Remove placebo .claude/hooks.json (echo-only 'safety systems') + duplicate .sh/.ts activation variants + 5 broken root hooks requiring nonexistent lib/
- Remove phantom skill-rules-v11.json (v8-labeled, 19/27 referenced skills nonexistent)
- Prune skill-rules.json to 20 rules that all resolve to real skills; brand_skill -> brand-voice; v12 stamp
- .mcp.json: drop claude-flow server; keep higgsfield multimodal connector
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
…tity contract - New CREATOR.md identity contract (root template); FRANK_DNA.md -> instances/frankx/CREATOR.md as the worked example; core is now adoptable by any creator - scripts/generate-stats.mjs: generated STATS.md + CI gates (stale counts, empty skills, phantom skill-rule targets); removed 4 more zero-byte skills it caught - CLAUDE.md rewritten for v12: honest counts via STATS.md, hook doctrine (telemetry-only), gstack/claude-flow as optional external integrations, no personal-machine mandates - README rewritten as a real front door: marketplace install first, npx skills add second, honest generated counts, related-projects trinity, live CI badge, no insider doctrine - AGENTS.md / GROK.md / QUICKSTART.md aligned; phantom commands (/ultrawork, /design-gods, /product-team-launch) removed from onboarding - /acos router: v7 header + '630+ skills' + claude-flow references corrected - New docs/ARCHITECTURE.md (the doc two other docs already linked), docs/ROADMAP.md (dated, honest), docs/VISION.md; CHANGELOG restarted at 12.0.0 - CONTRIBUTING.md: DCO sign-off + relicensing grant (open-core legal rail), identity-neutrality rule, hook doctrine; Frank-DNA alignment requirement removed - USAGE_GUIDE (v5 manual) archived; docs/README version claim fixed (was 3.0.0) - package.json files field now ships .claude/ + plugin manifests (npm artifact was missing the entire product) Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
- Point 19 files' FRANK_DNA.md references at the CREATOR.md identity contract - hook command/skill: hooks/ KB paths -> content-hooks/ - CODEOWNERS: .claude/hooks.json -> hooks/hooks.json - Optional-integration banners on 7 claude-flow wrapper commands (claude-flow no longer bundled) - alignment-check.yml messaging genericized (identity contract, not Frank DNA) Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
… disk claude plugin validate rejects directory paths for 'agents' (CLI 2.1.178). generate-stats.mjs now syncs the 65-file list into plugin.json and the --check gate fails CI when it drifts. 'claude plugin validate .' passes. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
…s resolved Leaks & employer content: - Normalize ALL case/format variants of personal machine paths (11 more files: /c/Users/frank, C:\Users\Frank, /home/frankx, mangled -mnt-c- forms) - Move luminor-strategic-guidance, frankx-website-builder, template-monetization, newsletter-ops to instances/frankx/ (personal/employer-strategy content) - Genericize all employer career/compliance/restriction references (public repo must not document employer strategy) - Scrub God-99/SHARING.md doctrine from grok layer + install.sh; install.sh v12 banner + version (was v10/11 fossil echoing purged persona lines) - Leak gate rewritten: case-insensitive, both path separators, telemetry+attribution patterns; no longer false-positives on the Oracle-spec academic citation. Simulated locally: CLEAN Hook pipeline (was 2/3 dead code): - Wire session-start.js (SessionStart) — the only creator of _active.json; without it post-tool-track and stop-finalize never ran - Strip output-style injection from session-start (violated telemetry-only doctrine) - Unify trajectory dir across all 4 hooks: CLAUDE_PROJECT_DIR (project), not plugin install dir (split-brain bug) - Widen PostToolUse matcher to Write|Edit|Bash so stop-finalize scoring heuristics are reachable - gitignore .claude/trajectories/ (session telemetry must never be committed) Skill activation quality: - kwToRegex: keywords starting with non-word chars (/studio) now match — command triggers were 100% dead - Remove dead-code continue that skipped fileTriggers for keyword rules - Drop 16 over-broad keywords (write, track, design, plan...) that injected false IMPORTANT guidance into ordinary coding prompts — verified fixed CI & supply chain: - alignment-check: drop tj-actions/changed-files (CVE-2025-30066 exposure) for git diff -z with injection-safe read loop; require only name+description (62/65 agents failed the old capabilities/priority gate); banned-language now warn-only - Remove ecosystem-monitor.yml (enumerated private repo names publicly, auto-committed to main) Truth & legal: - Kebab-case 39 agent frontmatter names (dispatch identifiers had spaces/ampersands) - CREDITS: Apache-2.0 attribution for 11 vendored Anthropic skills; ARCHITECTURE license claim corrected; 8->7 MCP servers - CONTRIBUTING: fix DCO/relicensing conflation (MIT already permits sublicensing) - README identity claim softened to match reality; OpenCode install claim corrected - Remove committed runtime artifacts: memory.db, homunculus instincts, gsd leftovers - research skill: convert orphaned CLAUDE.md to proper SKILL.md with frontmatter - generate-stats: CRLF-tolerant staleness check Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
|
Important Review skippedToo many files! This PR contains 980 files, which is 830 over the limit of 150. To get a review, narrow the scope: Upgrade to a paid plan to raise the limit. ⚙️ Run configurationConfiguration used: defaults Review profile: CHILL Plan: Pro Run ID: ⛔ Files ignored due to path filters (60)
📒 Files selected for processing (980)
You can disable this status message by setting the Use the checkbox below for a quick retry:
✨ Finishing Touches🧪 Generate unit tests (beta)
|
![gemini-code-assist[bot]](https://avatars.githubusercontent.com/in/956858?s=60&v=4){kind=small}
There was a problem hiding this comment.
Code Review
This pull request reorganizes and standardizes the Agentic Creator OS configuration. It removes several unused agent and coordinator files, introduces a new marketplace configuration, updates file paths to use home directory shortcuts instead of absolute Windows paths, and standardizes agent names to match their filenames. The review feedback correctly identifies that while several agent names (such as coaching-program-design, mcp-server-advising, and nano-banana-image-generation) were updated, their descriptions still contain references to their old names. These references should be updated to maintain consistency across the configuration files.
Important
The consumer version of Gemini Code Assist on GitHub is being sunset. Starting June 18, 2026, new organization installations will be blocked, and all code review activity will officially cease on July 17, 2026.
For more details on the timeline and next steps, please review the Help Documentation.
| @@ -1,5 +1,5 @@ | |||
| --- | |||
| name: coaching-program-designer | |||
| name: coaching-program-design | |||
There was a problem hiding this comment.
The agent's name has been updated to coaching-program-design to match the filename, but the description on line 3 still contains multiple references to the old name coaching-program-designer (e.g., coaching-program-designer agent). Please update these references in the description to ensure consistency and prevent potential lookup or invocation failures.
| @@ -1,5 +1,5 @@ | |||
| --- | |||
| name: mcp-server-advisor | |||
| name: mcp-server-advising | |||
There was a problem hiding this comment.
The agent's name has been updated to mcp-server-advising to match the filename, but the description on line 3 still contains multiple references to the old name mcp-server-advisor (e.g., mcp-server-advisor agent). Please update these references in the description to ensure consistency and prevent potential lookup or invocation failures.
| @@ -1,5 +1,5 @@ | |||
| --- | |||
| name: nano-banana-image-creator | |||
| name: nano-banana-image-generation | |||
There was a problem hiding this comment.
The agent's name has been updated to nano-banana-image-generation to match the filename, but the description on line 3 still contains multiple references to the old name nano-banana-image-creator (e.g., nano-banana-image-creator agent). Please update these references in the description to ensure consistency and prevent potential lookup or invocation failures.
Owner
Author
Swarm triage (2026-07-15)Status: PR #33 (model-arena sync) was merged separately (green CI). After rebase of this v12 open-core PR:
Related capture: issue tracking unfinished ACOS agent ideas + README 4-section excellence. |
This was referenced Jul 15, 2026
Owner
Author
Hermes PR Review Swarm — 2026-07-17Verdict: REQUEST CHANGES (process) — CONFLICTING + mega-diff; do not merge as-is Register
Blockers
Required before re-review
GEO/SEO
Hermes pr-review-swarm · do-not-merge until conflict-free + staged review. |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What this is
The v12 open-core reset, produced from a 7-domain multi-agent audit + adversarial verification pass. Theme: make every claim true. 1,040 files changed, −254K lines.
Why
The audit found the product real (100+ skills, ~70 commands, ~65 agents, 459 unique cloners/14d) but every credibility and distribution layer broken:
schedule:keyWhat changed
instances/frankx/; employer references genericized.claude-plugin/marketplace.json+ component paths —/plugin marketplace add frankxai/agentic-creator-osnow works;claude plugin validate .passesSTATS.md+scripts/generate-stats.mjs --checkCI gates (stale counts, empty skills, phantom rules, plugin agent-list drift) + case-insensitive leak gate — simulated locally: cleanVerification
node scripts/generate-stats.mjs --check→ all gates pass (101 skills / 68 commands / 63 agents / 4 hooks)claude plugin validate .→ passedFollow-ups (tracked in docs/ROADMAP.md)
🤖 Generated with Claude Code