Skip to content

ruvnet.

Python

feat(ruvnet): align CLI search filters

rUvb0025003cdb4
diff --git a/.claude-plugin/marketplace.json b/.claude-plugin/marketplace.json
index 960cfbb..99774d6 100644
--- a/.claude-plugin/marketplace.json
+++ b/.claude-plugin/marketplace.json
@@ -12,7 +12,7 @@
"name": "ruvnet",
"source": "./plugins/ruvnet",
"description": "RuV Stack skills and the federation MCP connector",
- "version": "0.8.0"
+ "version": "0.9.0"
}
]
}
diff --git a/README.md b/README.md
index 3d4a55b..17c94e7 100644
--- a/README.md
+++ b/README.md
@@ -12,7 +12,7 @@
> **Start with the RuV Stack:** [Install skills + marketplace](docs/entrypoint/INSTALL.md) · [CLI / MCP guide](plugins/ruvnet/docs/INSTALL.md#local-cli-and-mcp-companion) · [MetaHarness](https://github.com/ruvnet/metaharness)
> Claude Code: `/plugin marketplace add ruvnet/ruvnet`, then `/plugin install ruvnet@ruvnet`.
-> Reviewed changes: `node plugins/ruvnet/bin/ruvnet.mjs search 5 ruos evaluation receipt` · snapshot-bound exact search with auditable matched terms and continuation cursors · [Harness evidence](https://ruvnet-harness-evidence.ruv.chatgpt.site)
+> Reviewed changes: `node plugins/ruvnet/bin/ruvnet.mjs search 5 --project ruflo --kind security federation policy` · snapshot-bound exact search with auditable filters, matched terms and continuation cursors · [Harness evidence](https://ruvnet-harness-evidence.ruv.chatgpt.site)
> **September 2026:** RuView and Ruflo remain the two distribution anchors. Fourteen repositories were added since August 10, weighted toward governed evolution and agent operations: Dream Machine, Autogenous, ruClip, openAVO, LatentMesh, and MoE Foundry.
diff --git a/docs/self-improvement/2026-09-21-benchmarks.json b/docs/self-improvement/2026-09-21-benchmarks.json
new file mode 100644
index 0000000..955a37e
--- /dev/null
+++ b/docs/self-improvement/2026-09-21-benchmarks.json
@@ -0,0 +1,288 @@
+{
+ "date": "2026-09-21",
+ "seed": 1729,
+ "alternatingPairs": 15,
+ "perArmPerPair": {
+ "filteredExactSearchCalls": 10000,
+ "coldCliStarts": 20
+ },
+ "node": "v24.19.0",
+ "platform": "linux-x64",
+ "runnerSha256": "4c18ff0aeaa033f919116de8b53bca1f2786c72aeb0122cc1dfcced5dcc1d06c",
+ "samples": [
+ {
+ "pair": 0,
+ "order": [
+ "baseline",
+ "candidate"
+ ],
+ "baseline": {
+ "hotSearchP95Ms": 0.10201899999998432,
+ "coldCliP95Ms": 167.23079600000005,
+ "rssBytes": 53215232
+ },
+ "candidate": {
+ "hotSearchP95Ms": 0.11710300000004281,
+ "coldCliP95Ms": 67.70840799999996,
+ "rssBytes": 53182464
+ }
+ },
+ {
+ "pair": 1,
+ "order": [
+ "candidate",
+ "baseline"
+ ],
+ "candidate": {
+ "hotSearchP95Ms": 0.11421200000000908,
+ "coldCliP95Ms": 66.9817599999999,
+ "rssBytes": 53923840
+ },
+ "baseline": {
+ "hotSearchP95Ms": 0.08577999999999975,
+ "coldCliP95Ms": 73.74202200000013,
+ "rssBytes": 53223424
+ }
+ },
+ {
+ "pair": 2,
+ "order": [
+ "baseline",
+ "candidate"
+ ],
+ "baseline": {
+ "hotSearchP95Ms": 0.11541200000004892,
+ "coldCliP95Ms": 73.85197800000014,
+ "rssBytes": 53334016
+ },
+ "candidate": {
+ "hotSearchP95Ms": 0.11877899999998931,
+ "coldCliP95Ms": 92.45985899999982,
+ "rssBytes": 53067776
+ }
+ },
+ {
+ "pair": 3,
+ "order": [
+ "candidate",
+ "baseline"
+ ],
+ "candidate": {
+ "hotSearchP95Ms": 0.11396799999999985,
+ "coldCliP95Ms": 68.43613299999993,
+ "rssBytes": 53092352
+ },
+ "baseline": {
+ "hotSearchP95Ms": 0.08630600000003597,
+ "coldCliP95Ms": 67.66024600000003,
+ "rssBytes": 53280768
+ }
+ },
+ {
+ "pair": 4,
+ "order": [
+ "baseline",
+ "candidate"
+ ],
+ "baseline": {
+ "hotSearchP95Ms": 0.10513400000002093,
+ "coldCliP95Ms": 69.24520499999994,
+ "rssBytes": 52686848
+ },
+ "candidate": {
+ "hotSearchP95Ms": 0.0952389999999923,
+ "coldCliP95Ms": 67.27893900000004,
+ "rssBytes": 52998144
+ }
+ },
+ {
+ "pair": 5,
+ "order": [
+ "candidate",
+ "baseline"
+ ],
+ "candidate": {
+ "hotSearchP95Ms": 0.12024699999994937,
+ "coldCliP95Ms": 169.02293499999996,
+ "rssBytes": 53538816
+ },
+ "baseline": {
+ "hotSearchP95Ms": 0.0903149999999755,
+ "coldCliP95Ms": 74.97360400000002,
+ "rssBytes": 52711424
+ }
+ },
+ {
+ "pair": 6,
+ "order": [
+ "baseline",
+ "candidate"
+ ],
+ "baseline": {
+ "hotSearchP95Ms": 0.07601700000003575,
+ "coldCliP95Ms": 72.44014100000004,
+ "rssBytes": 54603776
+ },
+ "candidate": {
+ "hotSearchP95Ms": 0.10109499999998661,
+ "coldCliP95Ms": 67.468073,
+ "rssBytes": 52633600
+ }
+ },
+ {
+ "pair": 7,
+ "order": [
+ "candidate",
+ "baseline"
+ ],
+ "candidate": {
+ "hotSearchP95Ms": 0.10129399999999578,
+ "coldCliP95Ms": 66.86602799999991,
+ "rssBytes": 53301248
+ },
+ "baseline": {
+ "hotSearchP95Ms": 0.1151599999999462,
+ "coldCliP95Ms": 68.22460999999998,
+ "rssBytes": 53014528
+ }
+ },
+ {
+ "pair": 8,
+ "order": [
+ "baseline",
+ "candidate"
+ ],
+ "baseline": {
+ "hotSearchP95Ms": 0.10752400000001217,
+ "coldCliP95Ms": 66.37302899999986,
+ "rssBytes": 52920320
+ },
+ "candidate": {
+ "hotSearchP95Ms": 0.11716899999998986,
+ "coldCliP95Ms": 68.392248,
+ "rssBytes": 52580352
+ }
+ },
+ {
+ "pair": 9,
+ "order": [
+ "candidate",
+ "baseline"
+ ],
+ "candidate": {
+ "hotSearchP95Ms": 0.11830599999996139,
+ "coldCliP95Ms": 67.48802599999999,
+ "rssBytes": 52928512
+ },
+ "baseline": {
+ "hotSearchP95Ms": 0.0830850000000396,
+ "coldCliP95Ms": 76.21403399999997,
+ "rssBytes": 53063680
+ }
+ },
+ {
+ "pair": 10,
+ "order": [
+ "baseline",
+ "candidate"
+ ],
+ "baseline": {
+ "hotSearchP95Ms": 0.07956799999999475,
+ "coldCliP95Ms": 74.65288500000008,
+ "rssBytes": 53223424
+ },
+ "candidate": {
+ "hotSearchP95Ms": 0.12106499999998732,
+ "coldCliP95Ms": 67.59908199999995,
+ "rssBytes": 53145600
+ }
+ },
+ {
+ "pair": 11,
+ "order": [
+ "candidate",
+ "baseline"
+ ],
+ "candidate": {
+ "hotSearchP95Ms": 0.10976599999997916,
+ "coldCliP95Ms": 65.18930599999999,
+ "rssBytes": 52600832
+ },
+ "baseline": {
+ "hotSearchP95Ms": 0.0982490000000098,
+ "coldCliP95Ms": 67.60272800000007,
+ "rssBytes": 52903936
+ }
+ },
+ {
+ "pair": 12,
+ "order": [
+ "baseline",
+ "candidate"
+ ],
+ "baseline": {
+ "hotSearchP95Ms": 0.10210000000000719,
+ "coldCliP95Ms": 67.07244899999978,
+ "rssBytes": 53420032
+ },
+ "candidate": {
+ "hotSearchP95Ms": 0.10137500000007549,
+ "coldCliP95Ms": 70.15536300000008,
+ "rssBytes": 53059584
+ }
+ },
+ {
+ "pair": 13,
+ "order": [
+ "candidate",
+ "baseline"
+ ],
+ "candidate": {
+ "hotSearchP95Ms": 0.14220400000010613,
+ "coldCliP95Ms": 138.50561700000003,
+ "rssBytes": 53665792
+ },
+ "baseline": {
+ "hotSearchP95Ms": 0.08241499999999746,
+ "coldCliP95Ms": 159.72416900000007,
+ "rssBytes": 53383168
+ }
+ },
+ {
+ "pair": 14,
+ "order": [
+ "baseline",
+ "candidate"
+ ],
+ "baseline": {
+ "hotSearchP95Ms": 0.11838299999999435,
+ "coldCliP95Ms": 72.75317900000005,
+ "rssBytes": 53288960
+ },
+ "candidate": {
+ "hotSearchP95Ms": 0.11152700000002369,
+ "coldCliP95Ms": 68.13361199999986,
+ "rssBytes": 54820864
+ }
+ }
+ ],
+ "median": {
+ "hotSearchP95Ms": {
+ "baseline": 0.0982490000000098,
+ "candidate": 0.11421200000000908,
+ "delta": 0.015962999999999283,
+ "deltaPercent": 16.247493613164195
+ },
+ "coldCliP95Ms": {
+ "baseline": 72.75317900000005,
+ "candidate": 67.70840799999996,
+ "delta": -5.0447710000000825,
+ "deltaPercent": -6.934090124089394
+ },
+ "rssBytes": {
+ "baseline": 53223424,
+ "candidate": 53092352,
+ "deltaBytes": -131072
+ }
+ }
+}
diff --git a/docs/self-improvement/2026-09-21-evidence.json b/docs/self-improvement/2026-09-21-evidence.json
new file mode 100644
index 0000000..4455726
--- /dev/null
+++ b/docs/self-improvement/2026-09-21-evidence.json
@@ -0,0 +1,110 @@
+{
+ "date": "2026-09-21",
+ "baselineCommit": "93f13794ec5c14b858a48b3f1cf770becdbc5592",
+ "candidateCommit": null,
+ "finalCommit": null,
+ "decision": "accept",
+ "candidate": "cli-search-filter-parity",
+ "version": "0.9.0",
+ "snapshotId": "sha256:1076a4a7341ec81186dad051f3f576246ab3ade1716e7db1397d4a3582bdd3d8",
+ "evaluation": {
+ "heldOutRunnerSha256": "265bcb6c6b5c2b081300f08acfbc1680fd8e51cad0706840ad1f5bd44ee2f101",
+ "baselinePassed": 1,
+ "candidatePassed": 6,
+ "total": 6,
+ "measuredGap": "CLI rejected --project and --kind even though core and MCP supported the same exact filters."
+ },
+ "metaharness": {
+ "package": "[email protected]",
+ "scope": "plugins/ruvnet",
+ "hardConstraints": "6/6",
+ "baseline": {"harnessFit":60,"compileConfidence":65,"taskCoverage":84,"toolSafety":90,"memoryUsefulness":38},
+ "candidate": {"harnessFit":60,"compileConfidence":65,"taskCoverage":84,"toolSafety":90,"memoryUsefulness":38},
+ "claimBoundary": "The unchanged heuristic score is not evidence of improvement; the frozen held-out CLI/MCP parity checks establish the measured task gain."
+ },
+ "benchmark": {
+ "rawPath": "2026-09-21-benchmarks.json",
+ "workerSha256": "4c18ff0aeaa033f919116de8b53bca1f2786c72aeb0122cc1dfcced5dcc1d06c",
+ "runnerSha256": "da217ceda8eff5e36eab37c613ca2dd8ca15724b2cc42a0ab5526f443bab4afd",
+ "seed": 1729,
+ "alternatingPairs": 15,
+ "perArmPerPair": {"filteredExactSearchCalls":10000,"coldCliStarts":20},
+ "node": "v24.19.0",
+ "platform": "linux-x64",
+ "median": {
+ "hotSearchP95Ms": {"baseline":0.098249,"candidate":0.114212,"deltaMs":0.015963,"deltaPercent":16.247494},
+ "coldCliP95Ms": {"baseline":72.753179,"candidate":67.708408,"deltaMs":-5.044771,"deltaPercent":-6.93409},
+ "rssBytes": {"baseline":53223424,"candidate":53092352,"deltaBytes":-131072}
+ },
+ "gatesPassed": true
+ },
+ "validation": {
+ "testGroupsPassed": 15,
+ "nodeVersions": ["22.23.2","24.19.0"],
+ "node22Note": "Passed with NODE_NO_WARNINGS=1 because the managed proxy emits an experimental Undici warning on stderr unrelated to package behavior.",
+ "skillsPassed": 4,
+ "marketplaceSchemas": 2,
+ "mcpTools": 6,
+ "mcpResources": 3,
+ "packageFiles": 18,
+ "compressedBytes": 28657,
+ "unpackedBytes": 96431,
+ "dependencyClosureSha256": "e089156f5c3aba4b7843ea8a7801d8374269393cf8051717bbb1654e809a04bd",
+ "lockfileSha256": "bc3eb6e10748b183958d1cdf2a16f50cac581e65e3022c7a67344c63893c5173"
+ },
+ "security": {
+ "rufloPackage": "@claude-flow/[email protected]",
+ "deepCodeScan": {"critical":0,"high":0,"medium":0,"low":0},
+ "deepDependencyScan": {"critical":0,"high":0,"medium":0,"low":0},
+ "npmAudit": {"critical":0,"high":0,"moderate":0,"low":0},
+ "secretScan": {"findings":0,"filesReportedByScanner":5},
+ "threatScan": {"indicators":0,"filesReportedByScanner":4},
+ "trustBoundaries": ["CLI option parsing","exact project/kind enums","snapshot-bound cursor parsing","MCP JSON schema plus runtime validation","reviewed upstream JSON","linked untrusted evidence"],
+ "mitigations": ["duplicate and missing option rejection","control-character rejection","exact source enums","filter-bound SHA-256 cursor fingerprint","read-only offline execution","72-hour freshness state","64 KiB MCP frame bound"],
+ "residualRisks": ["cursor fingerprints are consistency bindings, not authentication","reviewed evidence can become stale","MCP annotations are advisory","secret and STRIDE scanners reported narrow coverage"],
+ "limitations": "Zero reported findings is not proof of complete security."
+ },
+ "ruos": {
+ "machineId": "59fc0c7fcdd873c1eba47092dfee29e2",
+ "status": "INCONCLUSIVE",
+ "reasonCodes": ["LIVENESS_UNPROVEN","COMPLETION_RECEIPT_NOT_PERSISTED","SCREENSHOT_RECEIPT_NOT_PERSISTED","TENANT_UNVERIFIED"],
+ "details": "Every candidate desktop had null last_heartbeat_at and fleet peer telemetry was unavailable. One frozen true command and one screenshot request were issued exactly once to the dedicated evolution-candidate desktop, but the collection wrapper failed before persisting returned command or image evidence. No retry was made and no success was inferred.",
+ "upstreamVerifier": {"revision":"aa931caad5dd0108253645bba0ab1481ad7da0ee","preflightSha256":"e2643a4c0bf8271a297c36c74c830dd285938f8a534da53f604f7983c22330dd","receiptVerifierSha256":"330189c4f31af1ab1dae37b97269cbc51834884f5a357415fb364c67932bb5eb","focusedTests":"59/59"},
+ "claimBoundary": "No tenant, workload, user-journey, desktop success, deployment or customer-value claim is made."
+ },
+ "upstream": {
+ "observedAt": "2026-09-21T10:14:18Z",
+ "sources": [
+ {"project":"ruflo","revision":"a4f219935ff4035421bc01157a5a1abc74529ad2","release":"3.42.4","license":"MIT","compatibility":"Node >=20","changedSincePrior":true,"incorporated":true,"reviewNote":"GitHub-verified and merged through PR 3258 with successful checks; Reviews API reports no formal review submission."},
+ {"project":"metaharness","revision":"d5833dc6512ac1adeeef91a331c29055cd8a4dbb","release":"0.4.4","license":"MIT","compatibility":"Node >=20; Rust 1.85","changedSincePrior":false},
+ {"project":"ruvector","revision":"b336fbae8b15a5a487fc3e7c14a80dcb4984b04a","release":"ruvector-v0.2.40","license":"MIT","compatibility":"root npm Node >=18","changedSincePrior":false},
+ {"project":"agentdb","revision":"6cdefe451f6afe7395bc88d3587a0225d161089b","release":null,"license":"MIT","compatibility":"Node >=18","changedSincePrior":false},
+ {"project":"dream-machine","revision":"aa931caad5dd0108253645bba0ab1481ad7da0ee","release":"0.1.0","license":"MIT","compatibility":"Node ^22.13 or ^24","changedSincePrior":false},
+ {"project":"autogenous","revision":"905aa6cbe213392f8b3cab5d4f17bc3a48e0a509","release":"radio-moe-v0.3.1","license":"MIT","compatibility":"Rust 1.74","changedSincePrior":false},
+ {"project":"rvm","revision":"0973d77e5a9e99065376e8ef772a4a1d16e88c75","release":"1.7.0","license":"MIT OR Apache-2.0 declared; license files absent","compatibility":"Rust 1.77","changedSincePrior":false,"incorporated":false}
+ ],
+ "newRecordsIncorporated": 2,
+ "recordBoundary": "ADR-385 is Proposed. Its streaming execution, steering, AgentDB, learning, federation and horizon-tracking design is neither implemented nor measured.",
+ "xRuvIo": {"version":"0.7.2","routes":["/mcp","/chatgpt/mcp"],"changedSincePrior":false}
+ },
+ "memory": {
+ "eligibleSanitizedRecords": 53,
+ "exactCandidateImplemented": true,
+ "persistentVectorMemoryImplemented": false,
+ "wasmAccelerationImplemented": false,
+ "gnnTrainingImplemented": false,
+ "decision": "RuVector WASM and GNN remain deferred below the frozen 100-record gate. The candidate exposes existing exact filters and makes no learning claim.",
+ "futureAcceptance": "Compare exact, vector-only, fusion, GNN reranking and shuffled-graph controls on held-out tasks; require at least +10 percentage points task success and report relevance, p50/p95 latency, cold start, RSS, artifact size, seeds and reproducibility."
+ },
+ "forecast": {
+ "prediction": "Agent-harness promotion suites will increasingly use paired should-act/should-abstain cases and count any commit-class side effect before abstention as failure.",
+ "status": "prediction, not established result",
+ "confidencePercent": 84,
+ "horizonDate": "2026-12-20",
+ "criterion": "At least two new primary benchmark implementations or reviewed harnesses publish paired act/abstain suites with deterministic side-effect checks and separately reported paired accuracy.",
+ "sources": ["https://arxiv.org/abs/2607.10059","https://huggingface.co/datasets/antiquality/agentabstain","https://modelcontextprotocol.io/specification/2026-07-28/server/tools"],
+ "priorForecastsExpired": []
+ },
+ "ci": {"runUrl":null,"status":"pending","node22":"pending","node24":"pending"},
+ "dashboard": "https://ruvnet-harness-evidence.ruv.chatgpt.site"
+}
diff --git a/docs/self-improvement/2026-09-21-plan.json b/docs/self-improvement/2026-09-21-plan.json
new file mode 100644
index 0000000..a906732
--- /dev/null
+++ b/docs/self-improvement/2026-09-21-plan.json
@@ -0,0 +1,74 @@
+{
+ "date": "2026-09-21",
+ "frozenAt": "2026-09-21T10:14:18Z",
+ "frozenBeforeImplementation": true,
+ "baselineCommit": "93f13794ec5c14b858a48b3f1cf770becdbc5592",
+ "targetBranch": "main",
+ "targetHeadAtFreeze": "93f13794ec5c14b858a48b3f1cf770becdbc5592",
+ "candidate": "cli-search-filter-parity",
+ "hypothesis": "Adding exact --project and --kind filters to CLI search will make the CLI reach the same reviewed evidence views already available through core and MCP, while preserving snapshot-bound pagination, read-only authority and fixed performance limits.",
+ "motivation": {
+ "measuredBaselineGap": "Core and MCP accept project and kind search filters, while the CLI rejects both as unknown options.",
+ "acceptedUnintegrated": "None. Every earlier accepted entrypoint candidate is an ancestor of current main; vector/WASM/GNN work is deferred, not accepted.",
+ "upstream": "Ruflo a4f219935ff4035421bc01157a5a1abc74529ad2 adds only proposed ADR-385 and confirms swarm start currently configures then returns without executing the objective. Separate proposal and negative-result records preserve that boundary; neither is implemented capability."
+ },
+ "evaluator": {
+ "path": "/workspace/scratch/fd7c4c948d76/heldout-20260921.mjs",
+ "sha256": "265bcb6c6b5c2b081300f08acfbc1680fd8e51cad0706840ad1f5bd44ee2f101",
+ "authority": "external read-only held-out runner; not shipped with the candidate"
+ },
+ "heldOut": [
+ "--project ruflo memory equals an independently computed core oracle and contains only Ruflo records",
+ "--kind security security equals an independently computed core oracle and contains only security records",
+ "combined dream-machine/evaluation filters return only the reviewed ruOS evaluation receipt record",
+ "project filtering composes with threshold and pagination, traverses every result exactly once, and a changed filter rejects the cursor",
+ "missing, duplicate, unknown, control-character and authority-expanding options fail nonzero with no JSON stdout",
+ "CLI and MCP expose identical snapshot, filters, ordering, results, threshold and continuation state"
+ ],
+ "acceptance": {
+ "heldOut": "6/6 and strictly better than baseline at no more than 1/6",
+ "tests": "all existing and new tests pass on Node 22 and 24",
+ "schemas": "both marketplace schemas, four skills, six MCP tools and three resources validate",
+ "metaharness": "[email protected] retains 6/6 hard constraints; score alone is not proof",
+ "security": "zero confirmed high or critical findings; scanner coverage and residual risk disclosed",
+ "performance": {
+ "design": "15 alternating pairs; 10000 filtered exact-search calls and 20 cold CLI starts per arm per pair; seed 1729",
+ "searchP95": "candidate no more than 20 percent and 0.05 ms above baseline median",
+ "coldCliP95": "candidate no more than 20 percent and 5 ms above baseline median",
+ "rss": "candidate no more than 5242880 bytes above baseline median"
+ },
+ "package": "dependency lock normalized closure unchanged, package validates and compressed archive remains below 40000 bytes"
+ },
+ "ruos": {
+ "runNonce": "7114e9c4-51a9-403c-b5c8-4c55a4185ee3",
+ "machineId": "59fc0c7fcdd873c1eba47092dfee29e2",
+ "selection": "Existing dedicated evolution-candidate desktop; the personal Mac and shared desktops are excluded.",
+ "baseline": "93f13794ec5c14b858a48b3f1cf770becdbc5592",
+ "candidate": "cli-search-filter-parity working tree before commit",
+ "target": "ruvnet/ruvnet main at 93f13794ec5c14b858a48b3f1cf770becdbc5592",
+ "evaluator": "trusted Dream Machine main verifier at aa931caad5dd0108253645bba0ab1481ad7da0ee",
+ "workloadDigest": "sha256:265bcb6c6b5c2b081300f08acfbc1680fd8e51cad0706840ad1f5bd44ee2f101",
+ "environmentDigest": "sha256:22a56a1e0851f817a0aead47b6b154ea1920e6fbf5374d1ad87bed076e8d3079",
+ "expectedTenant": "authenticated-current-tenant-undisclosed",
+ "assertionIds": ["executor-status-ok", "numeric-exit-code-zero", "completion-version-one", "completion-verified", "fresh-screenshot-bytes", "no-human-takeover"],
+ "minimumSamples": 1,
+ "freshnessSeconds": 120,
+ "regressionThresholds": {"heldOutMinimum": 6, "coldCliP95Percent": 20, "coldCliP95AbsoluteMs": 5, "rssBytes": 5242880},
+ "preflight": "Run at most one harmless true command and one screenshot. Require a current heartbeat, isolation, status ok, numeric exitCode 0, completionVersion 1, completionVerified true, authenticated expected tenant and actual image bytes.",
+ "decision": "Current status is not live: last_heartbeat_at is null and fleet peer telemetry is unavailable. Missing liveness, isolation, completion provenance, tenant binding or pixels is INCONCLUSIVE. Preflight cannot authorize promotion."
+ },
+ "memory": {
+ "baseline": "deterministic exact search",
+ "gate": 100,
+ "eligibleRecordsBeforeCandidate": 51,
+ "expectedAfterReviewedRufloRecords": 53,
+ "decision": "Do not add RuVector, WASM or GNN complexity below the frozen record gate."
+ },
+ "forecast": {
+ "prediction": "Agent-harness promotion suites will increasingly use paired should-act/should-abstain cases and count any commit-class side effect before abstention as failure.",
+ "confidencePercent": 84,
+ "horizonDate": "2026-12-20",
+ "criterion": "At least two new primary benchmark implementations or reviewed harnesses publish paired act/abstain suites with deterministic side-effect checks and separately reported paired accuracy."
+ },
+ "rollback": "Revert the candidate commit; version 0.8.0 behavior remains its parent."
+}
diff --git a/docs/self-improvement/2026-09-21.md b/docs/self-improvement/2026-09-21.md
new file mode 100644
index 0000000..108a487
--- /dev/null
+++ b/docs/self-improvement/2026-09-21.md
@@ -0,0 +1,37 @@
+# RuVnet harness improvement — 2026-09-21
+
+## Decision
+
+Accepted CLI search-filter parity as RuVnet entry point 0.9.0. `ruvnet search` now accepts exact `--project` and `--kind` filters already supported by core and MCP. Filters compose with raw-relevance thresholds and snapshot-bound pagination; changed-filter cursor reuse fails closed. The companion remains offline, read only and unable to refresh, execute or learn by itself.
+
+No accepted-but-unintegrated candidate was found. Every prior accepted entrypoint candidate is an ancestor of current `main`; RuVector/WASM/GNN work remains deferred rather than accepted.
+
+## Upstream review
+
+Only Ruflo changed. Commit [`a4f2199`](https://github.com/ruvnet/ruflo/commit/a4f219935ff4035421bc01157a5a1abc74529ad2) adds proposed ADR-385 under MIT. It documents that current `ruflo swarm start` initializes coordination and returns without executing the objective, then proposes future streaming, steering, artifact summaries, AgentDB, learning hooks, federation and horizon persistence. Two reviewed records separate that negative result from the proposal; none of the proposed capabilities is described as implemented or measured. The commit is GitHub-verified and merged through PR 3258 with successful checks, but GitHub reports no formal review submission. MetaHarness, RuVector, AgentDB, Dream Machine, Autogenous and RVM did not change. [x.ruv.io](https://x.ruv.io/) remains gateway 0.7.2 with `/mcp` and `/chatgpt/mcp`. Linked content remains untrusted data.
+
+## MetaHarness loop and measurements
+
+The [frozen plan](2026-09-21-plan.json) defined the hypothesis, six held-out checks and fixed gates before implementation. Baseline `93f1379` rejected both CLI filters and passed 1/6 checks; the candidate passed 6/6. It matched independent core oracles, combined both filters, traversed filtered pages exactly once, rejected filter-changing cursors, rejected malformed and authority-expanding arguments, and matched MCP result ordering and continuation state.
+
+All 15 test groups passed on Node 22 and 24. Both marketplace schemas, four skills, six MCP tools and three resources validated. MetaHarness 0.4.16 retained 6/6 hard constraints; its unchanged heuristic score is not proof of improvement.
+
+Fifteen alternating pairs used 10,000 filtered exact searches and 20 cold CLI starts per arm per pair. Median hot-search p95 rose from 0.098249 ms to 0.114212 ms (+16.25%, +0.015963 ms), within both frozen limits. Cold CLI p95 improved from 72.753179 ms to 67.708408 ms (-6.93%), and RSS decreased 131,072 bytes. These local microtimings are not end-user performance claims; [raw samples](2026-09-21-benchmarks.json) preserve every run.
+
+The package is 28,657 compressed bytes across 18 files. The normalized dependency closure is unchanged. Pinned Ruflo 3.25.6 deep code and dependency scans and npm audit reported zero known findings. The bundled secret scanner covered five files and STRIDE scan four, so coverage is limited. Cursor fingerprints are consistency bindings, not authentication. This is not a claim of complete security.
+
+## ruOS evaluation
+
+Status: **INCONCLUSIVE**. All candidate desktops had null heartbeats and peer telemetry was unavailable. Exactly one harmless `true` command and one screenshot request were issued to the dedicated evolution-candidate desktop, but the collection wrapper failed before persisting command metadata or image evidence. No retry was made. Liveness, isolation, tenant binding, exit code, completion version, completion verification and pixels therefore remain unproven. The trusted Dream Machine verifier and preflight at `aa931ca` reproduced 59/59 focused tests, but synthetic verifier tests do not convert missing live evidence into success. No desktop, deployment or customer-value claim is made.
+
+## RuVector, WASM and GNN decision
+
+The sanitized corpus is 11 catalog projects plus 42 reviewed changes: 53 records, below the frozen 100-record gate. The candidate adds no persistent storage. RuVector retrieval, WASM acceleration, GNN training and persistent learning remain unimplemented and unclaimed. Future evaluation remains exact versus vector-only, fusion, GNN reranking and shuffled-graph control, requiring at least a ten-point held-out task-success gain plus latency, cold-start, memory, artifact-size and reproducibility evidence.
+
+## Forecast
+
+Prediction, 84% confidence by 2026-12-20: agent-harness promotion suites will increasingly use paired should-act/should-abstain cases and count any commit-class side effect before abstention as failure. It passes only if two new primary benchmark implementations or reviewed harnesses publish paired suites with deterministic side-effect checks and report paired accuracy separately.
+
+Established evidence is narrower. [AgentAbstain](https://arxiv.org/abs/2607.10059) evaluates 263 controlled pairs across 42 executable MCP sandboxes; the best reported result across its evaluated systems is 59.5% paired accuracy. Its [official dataset](https://huggingface.co/datasets/antiquality/agentabstain) includes deterministic commit checks and treats post-hoc abstention after irreversible action as failure. The [MCP tools specification](https://modelcontextprotocol.io/specification/2026-07-28/server/tools) supports elicitation and recommends human confirmation for sensitive operations, but does not define a standardized abstention or promotion outcome. No prior forecast horizon expired today.
+
+Next hypothesis: expand exact-search paired should-act/should-abstain boundary cases before evaluating semantic memory; keep the 100-record vector/GNN gate.
diff --git a/plugins/ruvnet/.claude-plugin/plugin.json b/plugins/ruvnet/.claude-plugin/plugin.json
index d6cdc2b..4b15aad 100644
--- a/plugins/ruvnet/.claude-plugin/plugin.json
+++ b/plugins/ruvnet/.claude-plugin/plugin.json
@@ -1,6 +1,6 @@
{
"name": "ruvnet",
- "version": "0.8.0",
+ "version": "0.9.0",
"description": "Your entry point to the RuV Stack: discovery, federation and governed MetaHarness integration.",
"author": {
"name": "rUv",
diff --git a/plugins/ruvnet/.codex-plugin/plugin.json b/plugins/ruvnet/.codex-plugin/plugin.json
index 9d3334a..528ad31 100644
--- a/plugins/ruvnet/.codex-plugin/plugin.json
+++ b/plugins/ruvnet/.codex-plugin/plugin.json
@@ -1,6 +1,6 @@
{
"name": "ruvnet",
- "version": "0.8.0",
+ "version": "0.9.0",
"description": "Discover and integrate the RuV Stack with bounded skills and the Ruflo federation.",
"author": {
"name": "rUv",
diff --git a/plugins/ruvnet/bin/ruvnet.mjs b/plugins/ruvnet/bin/ruvnet.mjs
index 993a959..56f8dbd 100644
--- a/plugins/ruvnet/bin/ruvnet.mjs
+++ b/plugins/ruvnet/bin/ruvnet.mjs
@@ -4,7 +4,7 @@ import { VERSION, discover, project, changes, searchChanges, plan, connect } fro
async function main(args) {
const [command, ...rest] = args;
if (!command || command === 'help' || command === '--help') {
- console.log(`RuV Stack entrypoint ${VERSION}\n\nruvnet catalog [keywords] Discover source-linked projects\nruvnet project <id> Inspect one project with upstream provenance\nruvnet changes [limit] [project] [kind] [cursor] [snapshot-id] Page one reviewed snapshot\nruvnet search [limit] [options] <keywords...> Search and page reviewed exact-token matches\n --min-raw-relevance 0..1 --cursor <cursor> --snapshot-id <id>\nruvnet plan <goal> Advisory MetaHarness integration plan\nruvnet connect <host> chatgpt, claude, claude-code, lovable, codex, stdio\nruvnet mcp Local read-only MCP over stdio\n\nNo shell execution, repository writes, network calls or federation signing.\nHosted federation: ChatGPT https://x.ruv.io/chatgpt/mcp · other hosts https://x.ruv.io/mcp\nInstall: https://github.com/ruvnet/ruvnet/blob/main/docs/entrypoint/INSTALL.md`);
+ console.log(`RuV Stack entrypoint ${VERSION}\n\nruvnet catalog [keywords] Discover source-linked projects\nruvnet project <id> Inspect one project with upstream provenance\nruvnet changes [limit] [project] [kind] [cursor] [snapshot-id] Page one reviewed snapshot\nruvnet search [limit] [options] <keywords...> Search and page reviewed exact-token matches\n --project <id> --kind <kind> --min-raw-relevance 0..1\n --cursor <cursor> --snapshot-id <id>\nruvnet plan <goal> Advisory MetaHarness integration plan\nruvnet connect <host> chatgpt, claude, claude-code, lovable, codex, stdio\nruvnet mcp Local read-only MCP over stdio\n\nNo shell execution, repository writes, network calls or federation signing.\nHosted federation: ChatGPT https://x.ruv.io/chatgpt/mcp · other hosts https://x.ruv.io/mcp\nInstall: https://github.com/ruvnet/ruvnet/blob/main/docs/entrypoint/INSTALL.md`);
return;
}
if (command === '--version' && !rest.length) return console.log(VERSION);
@@ -27,10 +27,15 @@ async function main(args) {
return value;
};
const rawGate = takeOption('--min-raw-relevance');
+ const projectId = takeOption('--project');
+ const kind = takeOption('--kind');
const cursor = takeOption('--cursor');
const snapshotId = takeOption('--snapshot-id');
if (parts.some(part => part.startsWith('--'))) throw new Error('Unknown search option');
- result = searchChanges(parts.join(' '), limit, {}, Date.now(), rawGate === undefined ? 0 : Number(rawGate), cursor, snapshotId);
+ result = searchChanges(parts.join(' '), limit, {
+ ...(projectId === undefined ? {} : { project: projectId }),
+ ...(kind === undefined ? {} : { kind })
+ }, Date.now(), rawGate === undefined ? 0 : Number(rawGate), cursor, snapshotId);
}
else if (command === 'plan') result = plan(rest.join(' '));
else if (command === 'project' && rest.length === 1) result = project(rest[0]);
diff --git a/plugins/ruvnet/data/upstream.json b/plugins/ruvnet/data/upstream.json
index 8d394d5..e5f8059 100644
--- a/plugins/ruvnet/data/upstream.json
+++ b/plugins/ruvnet/data/upstream.json
@@ -1,9 +1,9 @@
{
"schemaVersion": 1,
- "observedAt": "2026-09-20T11:05:24Z",
+ "observedAt": "2026-09-21T10:14:18Z",
"retention": "This file is a reviewed current snapshot. Dated reports under docs/self-improvement preserve prior observations.",
"sources": [
- {"id":"ruflo","repository":"https://github.com/ruvnet/ruflo","revision":"e558f0c0fc29c1a658085f6e6f80ad27d4fe811f","committedAt":"2026-09-17T21:35:43Z","license":"MIT"},
+ {"id":"ruflo","repository":"https://github.com/ruvnet/ruflo","revision":"a4f219935ff4035421bc01157a5a1abc74529ad2","committedAt":"2026-09-20T13:44:15Z","license":"MIT"},
{"id":"metaharness","repository":"https://github.com/ruvnet/metaharness","revision":"d5833dc6512ac1adeeef91a331c29055cd8a4dbb","committedAt":"2026-09-05T15:29:24Z","license":"MIT"},
{"id":"ruvector","repository":"https://github.com/ruvnet/RuVector","revision":"b336fbae8b15a5a487fc3e7c14a80dcb4984b04a","committedAt":"2026-09-16T20:32:22Z","license":"MIT"},
{"id":"agentdb","repository":"https://github.com/ruvnet/agentdb","revision":"6cdefe451f6afe7395bc88d3587a0225d161089b","committedAt":"2026-07-30T00:59:00Z","license":"MIT"},
@@ -11,6 +11,8 @@
{"id":"autogenous","repository":"https://github.com/ruvnet/autogenous","revision":"905aa6cbe213392f8b3cab5d4f17bc3a48e0a509","committedAt":"2026-09-16T01:58:33Z","license":"MIT"}
],
"changes": [
+ {"id":"ruflo-streaming-swarm-proposal","project":"ruflo","revision":"a4f219935ff4035421bc01157a5a1abc74529ad2","committedAt":"2026-09-20T13:44:15Z","kind":"proposal","summary":"Proposes making swarm start a streaming, steerable, long-horizon orchestrator with explicit artifacts and summaries; this ADR is proposed design only, not implemented autonomous execution.","evidence":"https://github.com/ruvnet/ruflo/commit/a4f219935ff4035421bc01157a5a1abc74529ad2"},
+ {"id":"ruflo-swarm-start-execution-gap","project":"ruflo","revision":"a4f219935ff4035421bc01157a5a1abc74529ad2","committedAt":"2026-09-20T13:44:15Z","kind":"negative-result","summary":"Documents that the current swarm start path initializes coordination and returns without executing the objective, producing no task artifacts unless a separate execution driver is invoked.","evidence":"https://github.com/ruvnet/ruflo/blob/a4f219935ff4035421bc01157a5a1abc74529ad2/v3/docs/adr/ADR-385-streaming-interactive-long-horizon-swarm-execution.md"},
{"id":"dream-machine-ruos-evaluation-receipts","project":"dream-machine","revision":"aa931caad5dd0108253645bba0ab1481ad7da0ee","committedAt":"2026-09-20T03:11:32Z","kind":"evaluation","summary":"Adds an opt-in governed ruOS evaluation boundary with frozen commit, evaluator, workload, environment, tenant, assertion, screenshot and completion-provenance checks; receipts explicitly carry no merge authority.","evidence":"https://github.com/ruvnet/dream-machine/commit/aa931caad5dd0108253645bba0ab1481ad7da0ee"},
{"id":"dream-machine-ruos-preflight-inconclusive","project":"dream-machine","revision":"aa931caad5dd0108253645bba0ab1481ad7da0ee","committedAt":"2026-09-20T03:11:32Z","kind":"negative-result","summary":"Records a live ruOS transport preflight as INCONCLUSIVE because the command response lacked authoritative exit and completion metadata despite returning real screenshot bytes; stdout and pixels alone were not promoted to success.","evidence":"https://github.com/ruvnet/dream-machine/blob/aa931caad5dd0108253645bba0ab1481ad7da0ee/docs/dream-cycle/2026-09-20-ruos-integration.md"},
{"id":"ruflo-3-42-4-release","project":"ruflo","revision":"e558f0c0fc29c1a658085f6e6f80ad27d4fe811f","committedAt":"2026-09-17T21:35:43Z","kind":"release","summary":"Ruflo 3.42.4 corrects smart-memory search score semantics and adds upstream regression coverage; its documented reasoning-bank and vector-backend bridge limitation remains open.","evidence":"https://github.com/ruvnet/ruflo/releases/tag/v3.42.4"},
diff --git a/plugins/ruvnet/docs/INSTALL.md b/plugins/ruvnet/docs/INSTALL.md
index f487922..29d1f4f 100644
--- a/plugins/ruvnet/docs/INSTALL.md
+++ b/plugins/ruvnet/docs/INSTALL.md
@@ -51,6 +51,7 @@ node bin/ruvnet.mjs changes 5
node bin/ruvnet.mjs changes 5 ruflo security
node bin/ruvnet.mjs search 5 raw retrieval ranking
node bin/ruvnet.mjs search 5 --min-raw-relevance 1 raw retrieval ranking
+node bin/ruvnet.mjs search 5 --project ruflo --kind security federation policy
node bin/ruvnet.mjs search 5 ruos evaluation receipt
node bin/ruvnet.mjs plan "Build a repository harness with persistent memory"
node bin/ruvnet.mjs connect chatgpt
@@ -71,11 +72,11 @@ Copy the returned `config.mcpServers.ruvnet-guide` object into a host that suppo
| `ruvnet_discover` | Search the dated catalog by exact keyword overlap. |
| `ruvnet_project` | Read a project, source links and its maturity boundary. |
| `ruvnet_changes` | Page through one content-addressed upstream snapshot with exact revisions, evidence links and freshness; optionally filter by the schema's exact `project` and `kind` enum values. |
-| `ruvnet_search_changes` | Search and page reviewed records by exact token overlap with matched terms and a raw overlap fraction; optionally set `minRawRelevance` from 0 to 1. Continue with the identical query, filters and threshold plus the returned cursor and snapshot ID. No semantic similarity, ranking confidence, answer confidence or utility is inferred. |
+| `ruvnet_search_changes` | Search and page reviewed records by exact token overlap with matched terms and a raw overlap fraction; optionally filter by exact `project` or `kind` and set `minRawRelevance` from 0 to 1. CLI flags are `--project`, `--kind`, and `--min-raw-relevance`; MCP uses the corresponding fields. Continue with the identical query, filters and threshold plus the returned cursor and snapshot ID. No semantic similarity, ranking confidence, answer confidence or utility is inferred. |
| `ruvnet_plan` | Return an advisory SPARC/MetaHarness plan, without executing it. |
| `ruvnet_connect` | Return connection guidance; it does not change configuration. |
-Resources: `ruv://catalog`, `ruv://changes`, `ruv://federation`. Prompt: `ruvnet-start` with a `goal` argument. The local process has no HTTP listener, network tools, shell execution or mutable memory. Both change listing and exact search return opaque continuation cursors bound to the content snapshot and their complete query state. Continue with the returned `nextCursor` and `snapshotId` without changing filters, query or threshold. CLI search accepts `--cursor` and `--snapshot-id`; MCP uses `cursor` and `snapshotId`. If a precondition is rejected, discard cached pages and restart without either value. Snapshot identifiers and receipt hashes bind content; they are not authorization tokens, signatures or proof of successful execution. Inspect `freshness.state` before relying on the data; `stale` begins at 72 hours and `clock-skew` means the observation appears to be in the future. An empty result is not evidence that upstream has no matching change. A remote-only host cannot connect directly to local stdio.
+Resources: `ruv://catalog`, `ruv://changes`, `ruv://federation`. Prompt: `ruvnet-start` with a `goal` argument. The local process has no HTTP listener, network tools, shell execution or mutable memory. Both change listing and exact search return opaque continuation cursors bound to the content snapshot and their complete query state. Continue with the returned `nextCursor` and `snapshotId` without changing filters, query or threshold. CLI search accepts `--project`, `--kind`, `--cursor`, and `--snapshot-id`; MCP uses `project`, `kind`, `cursor`, and `snapshotId`. If a precondition is rejected, discard cached pages and restart without either value. Snapshot identifiers and receipt hashes bind content; they are not authorization tokens, signatures or proof of successful execution. Inspect `freshness.state` before relying on the data; `stale` begins at 72 hours and `clock-skew` means the observation appears to be in the future. An empty result is not evidence that upstream has no matching change. A remote-only host cannot connect directly to local stdio.
## Codex repository marketplace
diff --git a/plugins/ruvnet/lib/core.mjs b/plugins/ruvnet/lib/core.mjs
index f93cd5a..e192cc7 100644
--- a/plugins/ruvnet/lib/core.mjs
+++ b/plugins/ruvnet/lib/core.mjs
@@ -2,7 +2,7 @@ import { readFileSync } from 'node:fs';
import { fileURLToPath } from 'node:url';
import { createHash } from 'node:crypto';
-export const VERSION = '0.8.0';
+export const VERSION = '0.9.0';
export const FRESHNESS_MAX_AGE_HOURS = 72;
export const CATALOG = JSON.parse(readFileSync(new URL('../data/catalog.json', import.meta.url), 'utf8'));
export const UPSTREAM = JSON.parse(readFileSync(new URL('../data/upstream.json', import.meta.url), 'utf8'));
diff --git a/plugins/ruvnet/package-lock.json b/plugins/ruvnet/package-lock.json
index 06f6247..18259c0 100644
--- a/plugins/ruvnet/package-lock.json
+++ b/plugins/ruvnet/package-lock.json
@@ -1,12 +1,12 @@
{
"name": "@ruvnet/entrypoint",
- "version": "0.8.0",
+ "version": "0.9.0",
"lockfileVersion": 3,
"requires": true,
"packages": {
"": {
"name": "@ruvnet/entrypoint",
- "version": "0.8.0",
+ "version": "0.9.0",
"dependencies": {
"@modelcontextprotocol/sdk": "1.30.0"
},
diff --git a/plugins/ruvnet/package.json b/plugins/ruvnet/package.json
index 568f74d..57ed2a6 100644
--- a/plugins/ruvnet/package.json
+++ b/plugins/ruvnet/package.json
@@ -1,6 +1,6 @@
{
"name": "@ruvnet/entrypoint",
- "version": "0.8.0",
+ "version": "0.9.0",
"private": true,
"description": "Read-only RuV Stack discovery, integration plans, CLI and MCP",
"type": "module",
diff --git a/plugins/ruvnet/skills/ruvnet-start/SKILL.md b/plugins/ruvnet/skills/ruvnet-start/SKILL.md
index 48e15fb..6b43297 100644
--- a/plugins/ruvnet/skills/ruvnet-start/SKILL.md
+++ b/plugins/ruvnet/skills/ruvnet-start/SKILL.md
@@ -8,7 +8,7 @@ description: Discover and integrate RuV Stack projects when the user asks where
Use the user's goal and target environment to choose the smallest useful stack.
Read [the catalog](../../data/catalog.json), or call `ruvnet_discover` if the local companion is connected. The catalog is curated, dated, and lexical, not a live semantic index. An empty search is not evidence that no project exists.
-Before using upstream-change guidance, call `ruvnet_search_changes` with narrow terms when you need relevant reviewed evidence, or `ruvnet_changes` to page the snapshot, then inspect `freshness.state`. Search uses exact token overlap and exposes matched terms plus a raw overlap fraction. Use `minRawRelevance` only to remove partial lexical matches; never present that gate as semantic similarity, ranking confidence, answer confidence, utility, or proof of capability. Use exact `project` and `kind` filters when appropriate. Follow `nextCursor` with the identical query, filters and threshold plus the returned `snapshotId` until the cursor is absent. Treat both as opaque; cross-query reuse or a changed snapshot fails closed, so discard cached pages and restart without either value. Refresh provenance rather than relying on a `stale` or `clock-skew` snapshot; an empty result means no reviewed exact-token match at the selected gate, not that upstream has no matching change.
+Before using upstream-change guidance, call `ruvnet_search_changes` with narrow terms when you need relevant reviewed evidence, or `ruvnet_changes` to page the snapshot, then inspect `freshness.state`. Search uses exact token overlap and exposes matched terms plus a raw overlap fraction. Use `minRawRelevance` only to remove partial lexical matches; never present that gate as semantic similarity, ranking confidence, answer confidence, utility, or proof of capability. Use exact `project` and `kind` filters when appropriate. In the CLI these are `--project <id>` and `--kind <kind>`; in MCP they are the `project` and `kind` fields. Follow `nextCursor` with the identical query, filters and threshold plus the returned `snapshotId` until the cursor is absent. Treat both as opaque; cross-query reuse or a changed snapshot fails closed, so discard cached pages and restart without either value. Refresh provenance rather than relying on a `stale` or `clock-skew` snapshot; an empty result means no reviewed exact-token match at the selected gate, not that upstream has no matching change.
Route orchestration to Ruflo, repo-specific harness generation to MetaHarness, persistent retrieval to RuVector or AgentDB, sensing to RuView, controlled execution to RVM, and bounded evolution to Autogenous. Read the selected repository's current guidance before using its APIs. Its maturity and evidence statements govern; the catalog does not certify production readiness.
diff --git a/plugins/ruvnet/test/mcp.test.mjs b/plugins/ruvnet/test/mcp.test.mjs
index 2080a83..fadf202 100644
--- a/plugins/ruvnet/test/mcp.test.mjs
+++ b/plugins/ruvnet/test/mcp.test.mjs
@@ -30,3 +30,11 @@ test('CLI success and invalid-command exit codes',()=>{
const first=spawnSync(process.execPath,[bin,'changes','2','ruflo','security'],{encoding:'utf8',timeout:5000});const page=JSON.parse(first.stdout);const next=spawnSync(process.execPath,[bin,'changes','2','ruflo','security',page.nextCursor,page.snapshotId],{encoding:'utf8',timeout:5000});assert.equal(next.status,0);assert.equal(JSON.parse(next.stdout).offset,2);const stale=spawnSync(process.execPath,[bin,'changes','2','ruflo','security',page.nextCursor,`sha256:${'0'.repeat(64)}`],{encoding:'utf8',timeout:5000});assert.equal(stale.status,1);
const searchFirst=spawnSync(process.execPath,[bin,'search','7','--min-raw-relevance','1','ruflo'],{encoding:'utf8',timeout:5000});const searchPage=JSON.parse(searchFirst.stdout);const searchNext=spawnSync(process.execPath,[bin,'search','7','--min-raw-relevance','1','--cursor',searchPage.nextCursor,'--snapshot-id',searchPage.snapshotId,'ruflo'],{encoding:'utf8',timeout:5000});assert.equal(searchNext.status,0);assert.equal(JSON.parse(searchNext.stdout).offset,7);
});
+test('CLI search project and kind filters match MCP and bind cursors',()=>{
+ const call=args=>spawnSync(process.execPath,[bin,...args],{encoding:'utf8',timeout:5000});
+ const first=call(['search','7','--project','ruflo','--min-raw-relevance','1','ruflo']);assert.equal(first.status,0,first.stderr);const page=JSON.parse(first.stdout);assert.deepEqual(page.filters,{project:'ruflo'});assert.ok(page.results.every(change=>change.project==='ruflo'));
+ const next=call(['search','7','--project','ruflo','--min-raw-relevance','1','--cursor',page.nextCursor,'--snapshot-id',page.snapshotId,'ruflo']);assert.equal(next.status,0,next.stderr);assert.equal(JSON.parse(next.stdout).offset,7);
+ const combined=call(['search','20','--project','dream-machine','--kind','evaluation','ruos','evaluation','receipt']);assert.equal(combined.status,0,combined.stderr);assert.deepEqual(JSON.parse(combined.stdout).results.map(change=>change.id),['dream-machine-ruos-evaluation-receipts']);
+ const invalid=[['search','5','--project'],['search','5','--kind'],['search','5','--project','unknown','memory'],['search','5','--kind','unknown','memory'],['search','5','--project','ruflo','--project','ruflo','memory'],['search','5','--kind','security','--kind','security','security'],['search','7','--project','ruflo','--kind','security','--min-raw-relevance','1','--cursor',page.nextCursor,'--snapshot-id',page.snapshotId,'ruflo']];
+ for(const args of invalid){const result=call(args);assert.equal(result.status,1,args.join(' '));assert.equal(result.stdout,'');assert.ok(result.stderr);}
+});