agentic-engineering-harness 0.8.2 → 0.8.4

package/docs/CERTIFICATION.md

@@ -0,0 +1,31 @@
+# AEH Certification Core
+
+AEH certification is a runtime-agnostic quality boundary. `CertificationCore` accepts a `CandidateRevision`, a frozen `CertificationPolicy`, and an independent deterministic `CertificationOracle`. It emits one of:
+
+```text
+ACCEPTED
+REPAIR_REQUIRED
+HUMAN_REQUIRED
+PARTIAL
+BLOCKED
+NOT_TESTED
+```
+
+Capability reports contain separate `contract` and `modelE2E` lanes. A deterministic contract PASS with a blocked or untested model lane is `PARTIAL`, never a full PASS. A campaign requiring model execution is accepted only when the provider receipt proves startup and the oracle verifies model-produced journey evidence.
+
+External actors can propose or repair a candidate, but their claimed verdicts never become acceptance evidence. Provider execution is behind `AgentProvider`; the initial `CodexAgentProvider` is a bootstrap adapter and is not part of certification policy or oracle semantics.
+
+The bootstrap lane packs the current checkout, installs that pack into a disposable fixture with lifecycle scripts disabled, and evaluates it with an argv-based deterministic command oracle:
+
+```bash
+aeh certification self-dogfood <fixture> \
+  --check node \
+  --check-args '["-e","process.exit(0)"]' \
+  --prompt "Run the fixture task"
+```
+
+The Codex adapter resolves installed CLI capabilities from `codex exec --help`, uses `-c model_reasoning_effort="high"` when the CLI does not expose a reasoning flag, and records requested/effective configuration plus the Codex thread identifier. It uses a temporary controlled `HOME`/`CODEX_HOME` containing only the copied Codex auth file and generated model configuration; the fixture does not receive the host HOME tree.
+
+The provider boundary filters ambient environment variables, rejects recursive certification, bounds output and time, and terminates process groups on timeout or output overflow. Direct execution does not enforce OS-level network isolation. Reports therefore record `networkPolicy.requested`, `networkPolicy.enforced`, and `networkPolicy.enforcement`; a policy requiring isolation blocks when the direct transport cannot enforce it.
+
+Reports from the CLI are persisted under `.aeh-test-results/certification/<certification-id>.json` with the packed candidate SHA-256 and package identity. Temporary candidate paths remain diagnostic only.

package/docs/CONTEXT_EFFICIENCY.md

@@ -50,6 +50,36 @@ Budgets are role- and phase-aware. New projects default to `enforce`; `observe`
 
 Envelopes carry operation, agent, phase, budget, fragment projections, allowed retrieval IDs and a provenance hash. Large bodies are persisted as `.harness/context/<operation>/...raw`; envelopes contain references and compact content, not lifecycle authority.
 
+### Informational fast path
+
+Repository-grounded explanations use the same separation without creating an
+engineering operation. `aeh_informational_context` selects a bounded set of
+repository files and returns claims, summaries, direct repository-relative
+source refs, provenance and payload estimates. It does not write evidence,
+telemetry, operations, contracts or other `.harness` artifacts. Source bodies
+are not placed in `structuredContent` or replayed in `human`. The lead can call
+`aeh_informational_evidence` for one explicit
+`repo://...#sha256=<chunk>&file-sha256=<file>&range=<start>-<end>` reference
+when a summary is insufficient. A later bounded range can be requested by
+adding `&read=<start>-<end>` while retaining the original selected chunk and
+file identity. Retrieval verifies the whole file with a streaming hash and
+the selected chunk, retains only bounded returned bytes, and carries the
+requested excerpt once while structured metadata remains compact. Stale,
+forged, traversal and repository-escape references fail explicitly. Legacy
+range-only refs remain readable but cannot authorize a later range.
+
+Informational defaults are centralized at an 8K target, 12K soft limit and 15K
+exceptional limit, with eight source candidates, a 4KB per-source initial read
+limit and a 20KB total initial read limit. Projection and deduplication run
+before injection. Telemetry names the quantities it can actually measure:
+`rawEvidenceTokens`, `legacyPayloadTokens`, `projectedPayloadTokens`,
+`informationalPayloadTokens` and `duplicatePayloadTokensAvoided`. These are
+deterministic estimates of repository evidence and serialized payloads; they
+are not claims about actual lead-provider token usage. Source-code strings such
+as PASS/FAIL/WARN/security/uncertainty are summarized as implementation text,
+not promoted to verified runtime findings. Typed validation artifacts retain
+their separate authoritative evidence path.
+
 ## Runtime and troubleshooting
 
 Run `aeh doctor` after `aeh init --setup`. Doctor reports the gateway,

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "agentic-engineering-harness",
-  "version": "0.8.2",
+  "version": "0.8.4",
   "description": "OSS-first engineering harness for deterministic, spec-driven, issue-driven, audit-governed and orchestration-first multi-agent software delivery.",
   "type": "module",
   "bin": {
@@ -34,7 +34,7 @@
     "test:system-all": "npm run build && vitest run tests/system && AEH_RELIABILITY_SUITES=system AEH_RELIABILITY_REQUIRE_COMPLETE=1 tsx scripts/generate-reliability-report.ts",
     "reliability:report": "tsx scripts/generate-reliability-report.ts",
     "typecheck": "tsc -p tsconfig.json --noEmit",
-    "check": "npm run typecheck && npm test && npm run build",
+    "check": "npm run typecheck && npm run build && npm test",
     "release:check": "npm run check && npm pack --dry-run",
     "release:resolve": "node scripts/release-version.mjs"
   },

package/schemas/project.schema.json

@@ -69,6 +69,7 @@
     "security": { "type": "object", "properties": { "sandbox": { "type": "object", "properties": { "provider": { "type": "string" }, "required": { "type": "boolean" }, "image": { "type": "string" }, "imageDigest": { "type": "string" }, "network": { "type": "boolean" }, "extraArgs": { "type": "array", "items": { "type": "string" } }, "readOnlyRoot": { "type": "boolean" }, "ephemeralHome": { "type": "boolean" }, "noNewPrivileges": { "type": "boolean" }, "capDropAll": { "type": "boolean" }, "pidsLimit": { "type": "integer", "minimum": 1 }, "memory": { "type": "string" }, "cpus": { "type": "number", "exclusiveMinimum": 0 }, "tmpfs": { "type": "array", "items": { "type": "string" } }, "forceForRisks": { "type": "array", "items": { "enum": ["low", "medium", "high"] } }, "environmentAllowlist": { "type": "array", "items": { "type": "string" } }, "credentialEnvAllowlist": { "type": "array", "items": { "type": "string" } } } }, "tools": { "type": "array", "items": { "type": "string" } } } },
     "telemetry": { "type": "object" },
     "evals": { "type": "object", "properties": { "corpusDir": { "type": "string" }, "resultsDir": { "type": "string" }, "workspacesDir": { "type": "string" }, "defaultRuns": { "type": "integer", "minimum": 1 }, "confidenceLevel": { "type": "number", "exclusiveMinimum": 0, "exclusiveMaximum": 1 } } },
+    "certification": { "type": "object", "properties": { "enabled": { "type": "boolean" }, "policyPath": { "type": "string" }, "provider": { "type": "string" }, "fixtureRoot": { "type": "string" }, "requireSandbox": { "type": "boolean" }, "maxAttempts": { "type": "integer", "minimum": 0 }, "maxDurationMs": { "type": "integer", "minimum": 1 }, "maxCostUsd": { "type": "number", "minimum": 0 }, "maxTotalTokens": { "type": "integer", "minimum": 0 } } },
     "provenance": { "type": "object", "properties": { "outputDir": { "type": "string" }, "buildType": { "type": "string" }, "cosignKey": { "type": "string" }, "required": { "type": "boolean" }, "sbom": { "type": "object", "properties": { "required": { "type": "boolean" }, "command": { "type": "string" } } }, "signing": { "type": "object", "properties": { "required": { "type": "boolean" }, "key": { "type": "string" } } }, "verification": { "type": "object", "properties": { "required": { "type": "boolean" }, "publicKey": { "type": "string" } } } } }
   },
   "$defs": {

package/templates/project.yaml

@@ -202,6 +202,8 @@ context:
     default:
       inputTokens: 16000
     agents:
+      lead:
+        inputTokens: 12000
       explorer:
         inputTokens: 12000
       planner:
@@ -214,6 +216,14 @@ context:
         inputTokens: 14000
       operation-supervisor:
         inputTokens: 1500
+  informational:
+    targetTokens: 8000
+    softLimitTokens: 12000
+    exceptionalTokens: 15000
+    maxSources: 8
+    sourceSummaryTokens: 96
+    maxInitialBytesPerSource: 4000
+    maxInitialBytesTotal: 20000
   repositoryMap:
     enabled: true
     tokenBudget: 2000