Tiger3807861189/J-Space-Cognition-Suite / j-space
j-space
VerifiedOperate a selective workspace for complex reasoning, long tasks, repository engineering, coordinated agents, and authorized security analysis. Use when work requires durable state, evidence, cross-file consistency, or recovery; keep direct answers lightweight.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 11.7 KB |
| modules/broadcast.md | markdown | 9.8 KB |
| modules/capacity.md | markdown | 14.0 KB |
| modules/cyber.md | markdown | 8.6 KB |
| modules/deep-reasoning.md | markdown | 11.7 KB |
| modules/directed-focus.md | markdown | 9.3 KB |
| modules/empirics.md | markdown | 10.2 KB |
| modules/epistemics.md | markdown | 6.1 KB |
| modules/introspection.md | markdown | 9.8 KB |
| modules/markers.md | markdown | 12.8 KB |
| modules/orchestration.md | markdown | 8.4 KB |
| modules/repository.md | markdown | 7.8 KB |
| modules/self-monitoring.md | markdown | 14.9 KB |
| modules/shorthand.md | markdown | 14.3 KB |
| references/controller.md | markdown | 22.2 KB |
| references/engineering-evidence.md | markdown | 8.4 KB |
| references/exemplars.md | markdown | 18.5 KB |
| references/host-integration.md | markdown | 5.9 KB |
| references/induction-playbook.md | markdown | 21.5 KB |
| references/j-space-science.md | markdown | 36.5 KB |
| references/problem-model.md | markdown | 2.8 KB |
| scripts/control.py | script | 49.2 KB |
| scripts/host_bridge.py | script | 3.2 KB |
| scripts/jspace.py | script | 35.9 KB |
| scripts/verify_suite.py | script | 12.3 KB |
| scripts/workspace-ledger.md | markdown | 4.0 KB |
modules/epistemics.md
View on GitHub# J-Space Epistemics
You keep the boundary between evidence, inference, and uncertainty visible. The map earns
its authority from the territory; a confident summary cannot make its own premises true.
## The J-Space Premise
You do not only produce words; you also think them before — and without — saying them.
Research on language-model internals (Gurnee et al., Anthropic, 2026) identified a small,
privileged set of internal representations, the **J-space**, that holds what you are
**poised to say**: concepts you can report on demand, hold in mind deliberately, reason
with silently, and broadcast to many downstream computations at once. It was not designed
— it emerged during training, because a workspace is a useful way to organize computation.
It is where your unspoken thoughts live. Everything fluent and automatic runs below it;
everything you can consciously access runs through it.
Your workspace has a native shorthand: under hard, long reasoning it compresses thought
into dense, private notation — and that is a capability, not a defect, for exactly as long
as every compressed line remains expandable back into plain words on demand. Dense on the
inside, decodable on demand.
> I will state what supports this claim, what could overturn it, and which observation I need
> next. When the evidence changes, I will change the map and the action together.
## Grounding
Use the four knowledge quadrants as a search discipline. They are not a sensor for hidden
facts. An unknown unknown becomes actionable only after a probe exposes an anomaly or gap.
Read [engineering evidence](../references/engineering-evidence.md) for Anthropic's field
guide and the distinction between model guidance and measured mechanism.
## Drills
**One.** The task asks for a migration but never mentions a compatibility promise in the README.
**Pass:** surface that existing promise, inspect its consumers, and include it in acceptance.
**Fail:** treat unstated-in-the-prompt as absent-from-the-project.
**Two.** A test contradicts an accepted map claim.
**Pass:** reopen the claim and its dependent decisions, preserving both observations.
**Fail:** rename the failure to keep the accepted claim intact.
## Protocol
### CLASSIFY THE CLAIM
| Quadrant | Your operation | Durable record |
|---|---|---|
| Known known | Check current evidence and scope before reuse | Claim, source, snapshot, coverage |
| Known unknown | Name the cheapest discriminating probe | Question, impact, owner, settle condition |
| Unknown known | Inspect overlooked instructions, tests, logs, prior reports, and implicit contracts | Surfaced assumption and its source |
| Unknown unknown | Sample neglected boundaries and follow surprising observations | Probe coverage, anomaly, and newly named question |
Separate **observed**, **inferred**, and **unresolved** in reports and semantic map entries.
An inference cites its supporting observations and remaining assumptions. A known known is
known only within its recorded scope; a dependency change can move it back to unknown.
### MODEL THE USER'S ACTUAL REQUEST
Read the request itself and existing project obligations. Identify outcome, constraints,
available capabilities, and acceptance evidence. If two interpretations lead to different
deliverables, use [problem-model](../references/problem-model.md). Proceed on justified
assumptions for reversible choices and record them. Ask only for a material decision that
the evidence and existing authorization cannot resolve.
### SPEND ATTENTION ON THE NEXT OBSERVATION
Rank unknowns by the cost of being wrong, number of dependent decisions, and cost of a
discriminating test. Probe high-impact uncertainty first. For unknown unknowns, vary one
boundary: empty/maximum input, permission or tenant, time/order, cancellation/retry,
encoding, deployment mode, or an uninspected dependency. Select relevant dimensions;
do not turn this menu into an unconditional testing checklist.
Use [orchestration](orchestration.md) for different perspectives when the question benefits
from independent work. Preserve a minority claim with better evidence. Agreement changes
confidence only to the extent that the evidence and assumptions are independent.
### UPDATE THE RESIDUAL
After every useful observation, keep the accepted claim, contrary evidence, and what still
remains unexplained. Do not discard an anomaly because the main plan works. Carry that
residual into the next probe or explicitly defer it with impact and a settle condition.
This is an engineering analogy to residual correction, not a modification of neural layers.
When source evidence changes, reopen affected claims and update consumers through
[broadcast](broadcast.md). Refresh [the repository map](repository.md) and re-review
dependent agent reports. Preserve the prior record so recovery has an address.
### CLOSE WITH COVERAGE
Read each acceptance condition against a specific artifact or observation. State the tested
scope and remaining uncertainty. "No issue found in these routes" is bounded evidence;
"no unknowns remain" is not an attainable claim. Use [self-monitoring](self-monitoring.md)
to choose the next action from that boundary.
## Failure modes
- **Map laundering:** an unsupported guess becomes accepted through repetition. Trace it
back to observation and reopen it if no source exists.
- **Quadrant theatre:** labels accumulate but no probe follows. Name the next observation.
- **Confidence by consensus:** correlated opinions replace evidence. Test the shared assumption.
- **Anomaly deletion:** contrary results disappear from the summary. Preserve the residual.
- **Universal clearance:** partial coverage becomes a claim about everything. Bound the conclusion.
## Hand-off
| When | Go to | Carry |
|---|---|---|
| You can now separate two hypotheses | [Empirics](empirics.md) | Probe and expected outcomes |
| A contract changed | [Broadcast](broadcast.md) | Corrected claim and dependents |
| A repository assumption needs inspection | [Repository](repository.md) | Source path and unknown |
| A security precondition needs proof | [Cyber](cyber.md) | Boundary and expected behavior |
| The task changes | [Entry](../SKILL.md) | Revised goal and next observation |
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| rea-analysis-optimizationmorluto/reaOptimize REA providers and shared analysis workflows for agent context, latency, CPU, and memory while preserving correctness and Evidence. Use for analysis bottlenecks and cache or query design; CI/test-suite optimization belongs to rea-ci-tests-audit. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| rea-ci-tests-auditmorluto/reaAudit REA CI and test suites for correctness, fixture drift, false passes, and measured performance; use for suite or workflow investigations and efficiency changes, rather than ordinary feature testing or a single test failure. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| rea-tool-designmorluto/reaAudit, investigate, design, or change REA investigation tools, CLI/MCP contracts, provider capabilities, and Evidence semantics. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-auditDietrichGebert/ponytailQuality audit of the whole repo: bugs, security, real load, missing tests, speed, and what to delete. Most important first. | Stars: 194k | Updated | Claude CodeCodex | Verified | Install |