sheawinkler/ContextLattice

v3.25.0 Feature

This release adds 3 notable features for engineering teams evaluating rollout.

Published 7d AI Agents & Assistants

View tool

✓ No known CVEs patched

Read the diff → Tool health → What is this tool? →

✓ No known CVEs patched in this version

Topics

agent-framework agent-orchestration ai-infra ai-interface context-engineering context-management

+4 more

go long-horizon-agents long-horizon-intelligence rust

Summary

AI summary

Adds CLI commands for skill‑evolution candidate discovery, foundry handoff, and governed governance.

Changes in this release

Type	Severity	Summary	CVE
Feature
Feature	Medium	Adds `skill-evolution` CLI commands for reusable candidate discovery, retirement candidate handling, and Foundry handoff. Adds `skill-evolution` CLI commands for reusable candidate discovery, retirement candidate handling, and Foundry handoff. Source: llm_adapter@2026-07-19 Confidence: high	—
Feature	Medium	Introduces governed operator/enterprise lifecycle via `skill-evolution governance` CLI command. Introduces governed operator/enterprise lifecycle via `skill-evolution governance` CLI command. Source: llm_adapter@2026-07-19 Confidence: high	—
Feature	Medium	Enforces deterministic candidate and retirement identities, bounding server time to gateway rather than caller. Enforces deterministic candidate and retirement identities, bounding server time to gateway rather than caller. Source: llm_adapter@2026-07-19 Confidence: high	—
Feature	Medium	Keeps training evidence and holdout evaluation separate; no automatic activation, export, replacement, or retirement by public core. Keeps training evidence and holdout evaluation separate; no automatic activation, export, replacement, or retirement by public core. Source: llm_adapter@2026-07-19 Confidence: high	—
Feature	Medium	Performs atomic, replay‑safe Foundry handoff preserving workflow signature, prerequisites, rollback, side effects, constraints, and verification commands. Performs atomic, replay‑safe Foundry handoff preserving workflow signature, prerequisites, rollback, side effects, constraints, and verification commands. Source: llm_adapter@2026-07-19 Confidence: high	—
Feature	Medium	Specifies that the gateway never executes models, subprocesses, filesystem mutations, git operations, network adapters, merges, or pushes; external workers handle those actions. Specifies that the gateway never executes models, subprocesses, filesystem mutations, git operations, network adapters, merges, or pushes; external workers handle those actions. Source: llm_adapter@2026-07-19 Confidence: high	—
Feature	Medium	Defines rollback procedure: stop worker, deactivate schedules, submit exact receipt to rollback operation; preserves evidence and receipt history. Defines rollback procedure: stop worker, deactivate schedules, submit exact receipt to rollback operation; preserves evidence and receipt history. Source: llm_adapter@2026-07-19 Confidence: high	—

Full changelog

ContextLattice v3.25.0

Verified Skill Evolution

ContextLattice can now turn repeated verified work into a reviewable capability
without letting popularity masquerade as proof. The CLI remains the primary
interface. HTTP remains the integration fallback.

What changed

contextlattice_agent_tools skill-evolution discovers reusable workflow and
retirement candidates from bounded, evidence-bearing input.
Candidate evidence is resolved against persisted Utility Ledger outcomes and
agent-session verification events before a candidate can reach Skill Foundry.
Training evidence and holdout evaluation stay separate. No candidate is
activated, exported, replaced, or retired automatically by the public core.
Foundry handoff preserves the workflow signature, prerequisites, rollback,
side effects, platform constraints, and verification commands in one atomic,
replay-safe transaction.
Candidate and retirement identities are deterministic; server time is bound
by the gateway rather than trusted from callers.

Product boundary

The public core includes candidate discovery, evidence qualification, bounded
Skills Index projections, retirement protections, and explicit Foundry handoff.
It proposes. It does not activate.

Operator and Enterprise add workspace-scoped scheduling, governed evaluation,
approval, activation, monitoring, replacement, retirement, and rollback. Every
mutation requires explicit operator approval, generation matching, and an
idempotency key. Immutable receipts restore exact artifact identity.

The gateway never runs a model, subprocess, filesystem mutation, git operation,
network adapter, merge, or push. External workers execute operator-chosen work.

CLI

Public candidate discovery:

contextlattice_agent_tools skill-evolution reusable-candidate \
  --payload-file reusable-candidate.json --pretty

contextlattice_agent_tools skill-evolution retirement-candidate \
  --payload-file retirement-candidate.json --pretty

contextlattice_agent_tools skill-evolution foundry-handoff \
  --payload-file foundry-handoff.json --pretty

Governed Operator/Enterprise lifecycle:

contextlattice_agent_tools skill-evolution governance \
  --payload-file governance-request.json --pretty

Rollback

Stop the external discovery or evaluation worker, deactivate governed schedules,
and submit the exact activation, replacement, or retirement receipt to the
rollback operation. Preserve evidence and receipt history; never replace
rollback with deletion or mutable latest-state recovery.

View diff on GitHub

Weekly OSS security release digest.

The CVE patches and breaking changes that affected production tools this week. One email, every Sunday.

No spam, unsubscribe anytime.

Share this release

Share on X Share on Bluesky

Track sheawinkler/ContextLattice

Get notified when new releases ship.

About sheawinkler/ContextLattice

Private-by-default memory and context layer for agents with Go/Rust runtime, staged retrieval across fused data backends, and long-horizon context continuity.