v0.42.2 — MCP Definition Technical Preview

ShadowGraph 0.42.2 — Technical Preview

This patch publishes the accepted MCP-definition and mode-navigation corrections.
It preserves the existing tool capabilities and full/compact distinction. It is a
pre-release, not stable/GA. The package remains private: true and is not published
to npm; the v0.42.0 and v0.42.1 tags and existing assets remain unchanged.

Changes

• Qualify full-only tools referenced from compact-mode descriptions, including parameter and output-schema prose, with operator restart guidance. Preserve 35 default full tools, 16 compact tools, and 36 full tools with a configured verifier.
• Clarify tool selection, scoped read/write and review effects, record expansion parameters, maintenance and outcome behavior. Preserve callable names, validation constraints, output schema constraints and effect annotations.
• Correct compatibility documentation and identify the pinned local proxy gate as historical; do not change the proxy implementation.
• This metadata issue does not demonstrate an automatic-memory regression. Automatic-memory implementation and its accepted scope are unchanged; this patch does not repeat live memory validation.
• No new TDQS result is claimed by this release. A fresh Glama assessment, when available, is separate from release and deployment verification.

Install and upgrade

Use Node 22.5+ and PowerShell 7+ for the documented Windows Codex workflow:

npm install --global https://github.com/LiLara-AI/shadowgraph/releases/download/v0.42.2/shadowgraph-unified-plugin-0.42.2.tgz

Verify the downloaded tarball against SHA256SUMS.txt on the
GitHub Release.
The README covers the separate executor,
existing subscription and genuine owner consent. Installation alone does not enable
automatic memory. Stop active work and back up stores and the private integration
home before upgrading. Schema 7 is unchanged; changed source/executor/account
identity requires setup again, and older pending captures remain held rather than
being silently reassigned. See upgrade and rollback.

Scope and limitations

MCP names, validation rules, output schema constraints and effect annotations stay
compatible. Administrative capabilities remain full-only. Clearer prose adds metadata
bytes; this release is not a token-optimization or tool-choice benchmark claim.
The compatibility guide explains the supported workflows.

Automatic-memory behavior, project/account boundaries, sharing and recovery are
unchanged. The existing Windows Codex demonstration and limitations below remain
historical evidence, not a new live run. The Glama container provides credential-free
MCP; it cannot capture a user's Windows sessions. Publication, remote deployment and
a fresh TDQS assessment are distinct results; no score improvement is promised.

v0.42.1 — Windows Codex Technical Preview

ShadowGraph v0.42.1 — Technical Preview

Carry useful development experience into the next agent session through the installed Windows Codex terminal wrapper.

• Automatic capture of native command results and conversation text, with relevant experience supplied before later turns and fresh sessions.
• Completed answers and subsequent work proceed while memory processes. Accepted captures remain pending through delays/failures and recover through bounded idle/restart processing.
• Project-only defaults, explicit sharing/exclusions, corrections with history, and revocation that follows source dependencies.
• A clearer quick start, editable workflow diagram and preserved manual MCP/CLI/HTTP documentation.

Install

For the demonstrated Windows workflow, use PowerShell 7+, Node 22.5+ (24 recommended), and the separately installed Codex CLI 0.161.0 with an existing ChatGPT subscription and the explicit gpt-6.1-sol profile.

npm install --global https://github.com/LiLara-AI/shadowgraph/releases/download/v0.42.1/shadowgraph-unified-plugin-0.42.1.tgz

Follow the released README for executor discovery, genuine one-time owner consent, project enrollment and normal launch. Installing the package alone does not enable memory. ShadowGraph remains private: true and unpublished to npm; @openai/codex is a separate executor distribution.

The attached tarball's SHA-256 is:

f7f9ac3eac344695680b12254a6ddf20e7455aad3a9b9e87dfcf127da5776467  shadowgraph-unified-plugin-0.42.1.tgz

SHA256SUMS.txt is attached. The artifact contains 238 files. Its runtime source bytes match accepted candidate deab926eac3509a2fcd2aa4bf10593cd85a702b7; packaged documentation and the documentation inclusion list changed, so this tarball is not byte-identical to the earlier reviewed artifact.

Upgrade and limits

Stop active work and back up stores and the entire private integration home before upgrading. Schema 7 is retained. Changed source, executor or account identity requires setup again; older pending captures with different setup provenance are held, not silently reassigned. See upgrade and rollback.

This is a pre-release / Technical Preview, not stable/GA. The demonstrated automatic workflow observes only enrolled work launched with shadowgraph-codex run. Codex Desktop UI and a fresh Claude Desktop automatic chain are unproved. Existing Claude hooks retain separate host/activation limits. Each host turn is bounded to 120 seconds; extraction is eventual, and closing the wrapper leaves no permanent daemon. Capture does not cover every native tool-event type.

Canonical storage is local-first; selected capture/context data passes through the configured model provider. No universal recall, savings, superiority or human security audit claim is made. Glama exposes credential-free MCP, not capture of a user's local Windows sessions. No npm publication or standing production extraction is part of this release.

Codex guide · Full release notes · Report an issue

Never upload private stores, transcripts, credentials or review/evidence archives to a public issue.

ShadowGraph v0.41.0 — Technical Preview

Technical Preview / Early Access — not Beta, not Stable, not Production Ready.
Install from GitHub. The package remains private: true and is not published to npm.

ShadowGraph 0.41.0 keeps the decision-first graph and the scoped temporal memory from 0.40.0, and adds first-class explainable reconsideration built on the same condition evaluator review() already uses.

Install

npm install --global github:LiLara-AI/shadowgraph#v0.41.0
shadowgraph setup
shadowgraph doctor

First-class explainable reconsideration

review() answered one question — which decisions are due — and said nothing about the decisions it passed over. A caller could not tell "checked, and nothing fired" from "could not check." reconsider() answers both.

Verdict	When	Completeness
review_recommended	At least one relevant rule definitely fired	complete, or partial when something else could not be settled
unchanged	Evaluation was complete and nothing fired	always complete
manual_review	Nothing definitely fired, but something could not be settled	always partial

Each decision carries the evidence behind all three outcomes: triggeredRules, triggeredBy, groundedConditions, rulesNotEvaluated, contestedConditions, affectedAlternatives and factsConsidered.

Complete versus partial evaluation

unchanged is reported only with evaluationCompleteness: complete. That pairing is the contract's promise, and it is why an unevaluable condition can never be mistaken for a healthy decision.

The mixed case — one rule fires while another is unknown — is review_recommended and partial, so neither the breach nor the unevaluated condition hides the other. contestedConditions is reported apart from rulesNotEvaluated because the evaluator did reach a verdict there, but it withholds completeness all the same: a verdict resting on facts that disagree is exactly the silent pass the contract exists to prevent.

decisionId fails closed. An unknown id, an id belonging to another project, and a closed decision are each an error rather than an empty result. Key matching is literal: a near-miss key is unknown, never a match.

One shared three-valued condition evaluator

Reconsideration is a projection of the same pass review() runs, not a second evaluation. One pass, one evaluator, one set of three-valued verdicts — true, false, unknown — so the two routes cannot disagree about whether a stored rule fires, does not fire, or cannot be evaluated. Agreement is proven operator by operator from the canonical RULE_OPERATORS export rather than a second list that could drift.

Reconsideration across MCP, HTTP and CLI

shadowgraph_reconsider in both MCP modes, plus POST /reconsider and shadowgraph reconsider.

• Full mode: 28 tools (29 with an optional verifier configured)
• Compact mode: 14 tools — recommended, same full-fidelity stored graph

Counts are derived from the tool catalogue and asserted by npm run check:mcp, not restated by hand.

Runtime-authoritative MCP session provenance

sessionId was caller-owned in every direction: two unrelated clients could claim the same one, and a client that sent none left a null where the grouping belonged. The MCP server now mints one id per process and records it on every write it performs, so writes from one MCP session are attributable to that session by construction rather than by assertion.

A caller-supplied sessionId is still accepted — never an error, so no existing client breaks — but superseded, because asserted provenance does not outrank observed provenance. This covers the per-operation sessionId inside an applyMemoryPlan batch too. actor and client are untouched, CLI and HTTP callers are unaffected, and no migration is involved.

Benchmark v1.1 infrastructure and preserved evidence

The v1.1 harness, its frozen preregistration and amendment chain, adapter and probe infrastructure, and a sanitized evidence record are tracked so the methodology can be read and reproduced. "v1.1" names the benchmark methodology and candidate only — it is not a product release, and it carries no result. Raw benchmark output stays untracked, asserted by the test suite against both the Git index and the working tree.

Honest limits

• The package remains private: true and is NOT published to npm. This tag and release publish no package.
• No comparative benchmark result exists. The seven-arm comparative run measured zero arms because no common local/free LLM and embedding endpoint was available. No arm was scored, no arm was ranked, no arm is rank-eligible.
• The Phase-D decision-review diagnostic did NOT demonstrate a decision-quality benefit, and is not offered as evidence of one. The single acceptance run was explicitly non-scored and is valid for diagnostics only; it supports no comparison between arms.
• No claim that ShadowGraph outperformed the control or any other arm.
• No token, cost, latency or performance superiority claim.
• No ranking or best claim, or any equivalent overall-superiority wording.
• No human third-party security audit claim. An AI-assisted independent review passed on 2026-08-30 with no unresolved findings; it is a control, not a substitute. That gate remains open.
• Historical negative findings remain preserved in the repository as recorded — including that a substantial share of units the harness counted as measured did not produce a usable measurement under the frozen preregistration text. They were not rescored, reinterpreted, or removed.
• Semantic retrieval requires supplied embeddings or an explicitly configured endpoint; lexical fallback is never labelled semantic. Remote embeddings require an explicit privacy opt-in. Confidence calibration remains unresolved.

The only preregistered wording permitted for the current benchmark evidence:

Comparative benchmark infrastructure was executed, but no arm was measured because no common local/free LLM and embedding endpoint was available. No comparative performance, quality, token, cost, or 'best' claim is supported.

---

Full detail: RELEASE_NOTES.md · CHANGELOG.md · SECURITY.md