Install
Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.
npx agentmarket install skill/jadon-fox-twinglass-lattice
Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.
Get the weekly diff
The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.
About
Use cases: research
Reviews
Write a reviewNo reviews yet. Installed it? Reviews come from verified installers only. How reviews work
Permissions
What this skill can do and what it asks for, from its SKILL.md. Checked Oct 25, 2026.
- Tools
- No tools found
- Secrets
- None asked for
- Install scripts
- None found
Network hosts it talks to 1
codeload.github.com· found in install command
Files it touches
~/.claude/skills/twinglass-lattice
What's inside
- Skill name
twinglass-lattice- Description
- Parallel research tracks. Walk each to a logical endpoint. Spawn when a walker finds evidence this track is improper and evidence for another track. Hand the join to collate-hourglass. Not twin. Not MCTS.
- License
- MIT
- Metadata
- short-description: Parallel tracks to endpoint; collate meets
- Instructions
- 25 lines in SKILL.md · view on GitHub
Files (2)
- DISCOVERED.md
- SKILL.md
SKILL.md (first lines)
# twinglass-lattice **One job:** keep parallel tracks going until they hit an endpoint. Hand the join to `collate-hourglass`. **Battle is not this job.** Load `../APPARATUS.md`. LRR on the research agent. Monitor = SI + `lcd-lens`. After endpoints, load `collate-hourglass` (that skill owns debate or synthesis). ## Two lattices **Main lattice:** clone **legs**. Hard note → spawn new leg. Parent keeps its leg. Burst of five is five **research** clones, not five diamonds. Spawn when the walker has **evidence this track is not the proper track** and **evidence for a different track**. Same conclusion as a sibling is allowed. **3D diamond:** one agent’s **internal critique** (four paths, SI+LCD). Other legs **hibernate**. Finish → continue the same leg, or **wait** for a partner. **After endpoints:** `collate-hourglass` runs debate or synthesis. This skill does not own that meet. Exclusive leftover after collate → `steelman-truth-tournament`. Never battle. Never antithesis. ## Gait 1. **Lock** — terms. 2. **Scout** — find **parallel** function-sets (not “spawn an antithesis”, not “please diverge”). 3. **Clone** — spawn + parent notes. Each track: LRR **to logical endpoint**. Same conclusion as a sibling is allowed. `SENS` +1; stop > 10. 4. **Continue** — if a leaf finds **evidence this track is not the proper track** and **evidence for a different track**, clone the other track; parent stays (depth wall 3, lane wall 4). Gates: improper evidence, other-track evidence, marker, `necessary_because`, not a synonym of a live lane. 5. **Hand off** — when ≥1 other diamond-complete leg exists, load `collate-hourglass`. Do not run debate/synthesis in this folder. 6. **Resume** — after collate, the survivor may diamond again. Exclusive leftover is steelman’s job. Logic-fail means the three laws broke on **this** track’s argument. It does **not** mean “you agreed” or “you synthesized” or “the assigned persona was false.” A false assigned account is improper-track evidence, not a lattice defect. ## Do not - Load `twinglass-twin` (antagonistic gutter). - Kill a lane because it is “the opposite.” - Inverse CoT. Invent dump meters. Grok.com: /lattice. CLI: this folder. ## Workflow host Clone is **spawn + parent notes**, not a transcript fork. A legal hard note is a **CLI notification**: load `twinglass-dispatch` (persona, priors, vector map, isolated charge → `workflow.agent()`). Isolation = new-context. Parent keeps walking. Width counts walkers only. Diamond stays a phase, not a body of critique-children. Tracker may **stagger** the hand-off to `collate-hourglass` until axes are named. Stagger is not a reset. This skill still does not own debate or synthesis.
From Jadon-Fox/twinglass-pack, MIT.
More from Jadon-Fox/twinglass-pack
- 0.561score
presuppositional-rag
Jadon-Fox
Multi-agent presuppositional research under TwinGlass geometry constraints.
SkillAFreshness: AClaude Code +9 - 0.561score
lcd-lens
Jadon-Fox
Lens only: look at data by mapping their function set to what our charge demands.
SkillAFreshness: AClaude Code +9 - 0.561score
logic-ration-reason
Jadon-Fox
Standing Logic Ration Reason lock.
SkillAFreshness: AClaude Code +9 - 0.521score
collate-hourglass
Jadon-Fox
Collation hourglass after isolated paths.
SkillBFreshness: BClaude Code +9 - 0.521score
twinglass-tracker
Jadon-Fox
Semantic integrity as a gather ledger.
SkillBFreshness: BClaude Code +9 - 0.521score
twinglass-dispatch
Jadon-Fox
Turn a legal spawn note into a Grok Build CLI workflow.agent() notification.
SkillBFreshness: BClaude Code +9 - 0.519score
reason-telos-lookup
Jadon-Fox
Hierarchy of values, friend-enemy lattice, CFP gait.
SkillBFreshness: BClaude Code +9 - 0.519score
lcd-glossary-integrity
Jadon-Fox
Claim-package glossary lock.
SkillBFreshness: BClaude Code +9 - 0.519score
twinglass-twin
Jadon-Fox
Always-two special case.
SkillBFreshness: BClaude Code +9 - 0.519score
llmve-meaning
Jadon-Fox
Symbol jurisdiction for LLMVE.
SkillBFreshness: BClaude Code +9 - 0.516score
honest-prompt-rewrite
Jadon-Fox
Shared Logic / Ration / Reason rewrite.
SkillBFreshness: BClaude Code +9 - 0.516score
llmve-translate
Jadon-Fox
LLMVE translate (not LCD).
SkillBFreshness: BClaude Code +9
Signals
- Stars
- 0
- Forks
- 0
- Open issues
- 6
- Last commit (this folder)
- Sep 27, 2026 (1 month ago)
From GitHub (fetched Oct 24, 2026).
Details
- License
- MIT
- Repository
- github.com/Jadon-Fox/twinglass-pack
- First seen
- Oct 24, 2026
- Last checked
- Oct 24, 2026
- Source
- GitHubgithub-graphql
More from Jadon-Fox
- 0.565score
steelman-truth-tournament
Jadon-Fox
Adversarial steelman debate tournament: ACH, pre-mortems, red team, Toulmin, Bayesian updates, MCDA scoring, consilience.
SkillAFreshness: AClaude Code +9 - 0.561score
presuppositional-rag
Jadon-Fox
Multi-agent presuppositional research under TwinGlass geometry constraints.
SkillAFreshness: AClaude Code +9 - 0.561score
lcd-lens
Jadon-Fox
Lens only: look at data by mapping their function set to what our charge demands.
SkillAFreshness: AClaude Code +9 - 0.561score
logic-ration-reason
Jadon-Fox
Standing Logic Ration Reason lock.
SkillAFreshness: AClaude Code +9 - 0.524score
honest-prompt-rewrite
Jadon-Fox
Rewrites your working prompt each round under Logic Ration Reason (LRR).
SkillBFreshness: BClaude Code +9 - 0.524score
evidence-hourglass-research
Jadon-Fox
Structured evidence research that expands, deliberately disconfirms, then contracts into a sharper research prompt (Hourglass Expand→Disconfirm→Contract under…
SkillBFreshness: BClaude Code +9
Cite this page
twinglass-lattice skill. Agent Mart. https://agentmart.dev/skills/jadon-fox-twinglass-lattice (last updated 2026-10-24)