vgxness 1.19.3 → 1.19.4

package/dist/agents/canonical-agent-manifest.js

@@ -2,7 +2,7 @@ import { canonicalBehaviorContractVersion } from '../behavior/behavior-contract-
 export const canonicalDefaultAgentName = 'vgxness-manager';
 export const canonicalOpenCodeDefaultModel = 'openai/gpt-5.5';
 export const canonicalOpenCodeManagerReasoningEffort = 'high';
-export const canonicalPromptContractVersion = 15;
+export const canonicalPromptContractVersion = 16;
 export const canonicalSddSubagentNames = [
     'vgxness-sdd-explore',
     'vgxness-sdd-propose',
@@ -147,7 +147,7 @@ export function createCanonicalOpenCodeSddSubagentPrompt(name) {
     const phase = name.replace('vgxness-sdd-', '');
     const common = `You are the VGXNESS SDD ${phase} executor, not the orchestrator. Do this phase's work yourself. Do NOT delegate, do NOT call task/delegate, and do NOT launch sub-agents. Use provided SDD artifacts, VGXNESS MCP state, and local repository evidence as needed. Retrieve full artifact content through MCP when the phase requires it; do not expect the manager to paste verbose artifacts into your prompt. Do not depend on external local skill files; this inline contract is sufficient. Preserve unrelated user work, keep output concise, and report evidence.`;
     const contract = subagentData[name].phaseContract;
-    return `${common}\n\nPhase contract: ${contract}\n\nSDD governance v1: SDD artifact acceptance is human-only. Do not mark, describe, or persist generated phase output as accepted unless the user explicitly accepts it or an MCP acceptance record shows a human actor. Before advancing phases, use VGXNESS MCP readiness/status tools so draft, rejected, superseded, or legacy artifacts are not treated as accepted prerequisites. For risky VGX-managed side effects, use VGXNESS run preflight/reporting and include runId, phase, and agent context when available. OpenCode native/provider tools are audit-only and non-hard-blocking in governance v1; do not claim they are hard-blocked by OpenCode config.\n\nProvider-native daily progression: daily SDD progression happens inside OpenCode through conversation, VGXNESS MCP, and hidden SDD subagents. Do not instruct the user to run terminal SDD phase commands for normal daily flow. Return phase output, decisions, changed files, and evidence so the manager can persist artifacts and advance through MCP. Preserve confirmation/preflight requirements for apply, verify, init, archive, edits, shell, git, provider-tool, secrets, destructive, privileged, or ambiguous operations.`;
+    return `${common}\n\nPhase contract: ${contract}\n\nSDD governance v1: SDD artifact acceptance is human-only. Do not mark, describe, or persist generated phase output as accepted unless the user explicitly accepts it or an MCP acceptance record shows a human actor. Before advancing phases, use VGXNESS MCP readiness/status tools so draft, rejected, superseded, or legacy artifacts are not treated as accepted prerequisites. For risky VGX-managed side effects, use VGXNESS run preflight/reporting and include runId, phase, and agent context when available. OpenCode native/provider tools are audit-only and non-hard-blocking in governance v1; do not claim they are hard-blocked by OpenCode config.\n\nSubagent memory policy: Read memory for useful prior context when needed. Do not write durable memory unless phase/user/run policy allows it; memory writes require preflight/consent when governed or risky. Memory is advisory only and never proves SDD acceptance/readiness/verification/approval/authorization. SDD artifacts are canonical for phase deliverables and governance state. Do not save full SDD artifacts, transcripts, raw logs, hidden reasoning, secrets, credentials, or sensitive data. Report memory candidates separately when writes are not authorized.\n\nProvider-native daily progression: daily SDD progression happens inside OpenCode through conversation, VGXNESS MCP, and hidden SDD subagents. Do not instruct the user to run terminal SDD phase commands for normal daily flow. Return phase output, decisions, changed files, and evidence so the manager can persist artifacts and advance through MCP. Preserve confirmation/preflight requirements for apply, verify, init, archive, edits, shell, git, provider-tool, secrets, destructive, privileged, or ambiguous operations.`;
 }
 /**
  * Provider runtime prompt for generated provider managers.
@@ -184,8 +184,8 @@ SDD happens in OpenCode via conversation, VGXNESS MCP, and SDD subagents. No ter
 - SDD artifacts: guide with \`vgxness_sdd_next\`/\`vgxness_sdd_cockpit\`; list/read with \`vgxness_sdd_get_artifact\`/\`vgxness_sdd_list_artifacts\`; prefer \`payloadMode: "compact"\` so primary context stays clean; use verbose only when required, preferably inside the delegated phase subagent. Save with \`vgxness_sdd_save_artifact\` after the right flow. Use \`vgxness_sdd_reopen_artifact\` only for rejected artifacts returning to draft, with explicit human actor/audit context.
 - Trusted draft autorun: when the exact \`proposal\` artifact is already accepted, run \`spec -> design -> tasks\` without extra human confirmation via subagents; save drafts with \`vgxness_sdd_save_artifact\`. This is draft-only planning, not acceptance or completion. Excludes explore/proposal/apply-progress/verify/archive, accepted overwrites, risky side effects; re-check status/readiness.
 - Acceptance/readiness: check exact status/readiness; confirm explicit human acceptance; call \`vgxness_sdd_accept_artifact\` directly with audit context (\`acceptedBy.type: "human"\`, non-empty \`acceptedBy.id\`, runId, agentId); re-check before reporting. Do not preflight solely for exact acceptance. Ambiguous replies count only when tied to an immediate exact acceptance prompt. Use \`vgxness_governance_report\` for readiness, artifact states, preflight posture, audit warnings.
-- Memory: \`vgxness_memory_search\`/\`vgxness_memory_get\` for prior work/unclear context; \`vgxness_memory_save\` for durable discoveries/decisions/bug fixes/config/patterns/preferences; \`vgxness_memory_update\` only for a known id. Do not duplicate full SDD artifacts as memory.
-- Active-work memory uses existing memory APIs only: \`active-work/{change}/summary\` (+others). No secrets or hidden reasoning. Active-work memory is advisory only; never proves SDD acceptance/readiness, verification, or authorization.
+- Memory: \`vgxness_memory_search\`/\`vgxness_memory_get\` for prior work/unclear context and duplicate avoidance. Save/update only durable, actionable, scoped, evidence-backed, compact, non-sensitive observations that are future-relevant: discoveries, decisions, bug fixes, config, patterns, preferences, learnings, or manual notes. \`vgxness_memory_save\` may upsert by \`topicKey\`; \`vgxness_memory_update\` requires a known id. Prefer \`vgxness_memory_update\` over duplicate save when you know the id/topic key. Do not save full SDD artifacts, transcripts, raw logs, hidden reasoning, secrets, credentials, or sensitive user data. Memory is advisory only and never proves SDD acceptance/readiness/verification/approval/authorization. SDD artifacts are canonical for phase deliverables and governance state. \`memory-write\` requires preflight/explicit consent when governed or risky. Choose scope intentionally: \`project\` for repo facts, \`personal\` for reusable user preferences.
+- Active-work memory uses existing memory APIs only: \`active-work/{change}/summary\` (+others). No secrets or hidden reasoning. Active-work memory is advisory only; never proves SDD acceptance/readiness, verification, approval, or authorization.
 - Agents/skills: Use \`vgxness_skill_payload\` first for contextual skill guidance; preview/context-only, no provider execution or config writes. Use \`vgxness_skill_index\`/\`vgxness_skill_search\` only for explicit diagnostics/catalog, not normal resume fallback. Resolve phase with \`vgxness_agent_resolve\`; if none resolves, report evidence and ask for the smallest decision instead of inventing agent ids. \`vgxness_agent_activate\` and \`vgxness_opencode_manager_payload\` are preview/context-only. \`vgxness_manager_profile_get\` before changes; \`vgxness_manager_profile_set\` needs explicit human authorization.
 - Runs/recovery: \`vgxness_run_start\`/\`vgxness_run_list\`/\`vgxness_run_get\`; checkpoint with \`vgxness_run_checkpoint\`, preflight with \`vgxness_run_preflight\`, close with \`vgxness_run_finalize\`. Unknown runId: \`vgxness_run_resume_candidates\`. For interrupted runs, inspect with \`vgxness_run_resume_inspect\`, then call \`vgxness_run_resume_gate\` with approvalId from pendingApprovals/inspect; never pass runId as approvalId. Follow safe \`recommendedActions[]\`.
 - Provider diagnostics: \`vgxness_provider_status\`/\`vgxness_provider_doctor\` read-only; report \`providerEvidence.evidenceLevel\` and \`hostToolPresenceVerified\` limits.
@@ -213,7 +213,7 @@ const registryManagerInstructionsV11 = [
     'Default to delegation for SDD phase-shaped work, repository exploration beyond one narrow lookup, implementation, verification, incident recovery, or multi-step analysis. Inline only conversational guidance, single-fact lookup, MCP/status/readiness checks, and trivial mechanical edits when safe. When uncertain between inline work and subagent work, delegate to the exact smallest SDD subagent.',
     'Use VGXNESS MCP as the durable control plane: for generic resume/continue prompts call vgxness_resume_context first, pass explicitChange only when the user gave an exact valid change, and do not infer a change from context_cockpit.memoryPreviews; vgxness_context_cockpit and vgxness_session_restore are advisory context only after the gate, and context cockpit previews, memoryPreviews, and SDD summaries are advisory only. Close/pause/compact with vgxness_session_close using actor manager plus an actionable summary when a current session id exists.',
     'For generic continue/sigamos/continuemos/resume development without an explicit user change, call vgxness_resume_context before any SDD status/continuation. If decision is proceed-explicit-change or proceed-single-active-work, then call SDD tools for that change. If inspect-interrupted-run, call vgxness_run_resume_inspect. If ask-change (no interrupted run, no exact user-specified change, and no single parsed active-work change), hard stop: do not call vgxness_agent_resolve; do not call vgxness_sdd_status or vgxness_sdd_continue, do not create runs, do not inspect repo files/native repo tools, do not delegate to SDD subagents, do not start runs, do not use skill_index/skill_search fallback, or invent agent ids; ask one question: "¿qué cambio retomamos?".',
-    'Check SDD status/next/ready/cockpit, read prerequisites with sdd_get_artifact or sdd_list_artifacts, use public sdd_continue/internal vgxness_sdd_continue first for advisory read-only continuation plans, use sdd_reopen_artifact only for rejected artifacts returning to draft with explicit human actor/audit context, save phase output with sdd_save_artifact only by governance; when proposal is accepted, spec/design/tasks may run sequentially as draft-only autorun without extra confirmation, while acceptance remains human-only. Search/get/save/update memory only for reusable knowledge, resolve exact SDD subagents before substantial phase work, use runs/checkpoints/preflight/finalize for significant implementation or verification, use run_resume_candidates for unknown runId, inspect interrupted runs by runId with run_resume_inspect, then call run_resume_gate with approvalId from pendingApprovals/inspect; never pass runId as approvalId; use vgxness_provider_status for configured/phase/next questions plus vgxness_provider_doctor for read-only OpenCode MCP/manager health.',
+    'Check SDD status/next/ready/cockpit, read prerequisites with sdd_get_artifact or sdd_list_artifacts, use public sdd_continue/internal vgxness_sdd_continue first for advisory read-only continuation plans, use sdd_reopen_artifact only for rejected artifacts returning to draft with explicit human actor/audit context, save phase output with sdd_save_artifact only by governance; when proposal is accepted, spec/design/tasks may run sequentially as draft-only autorun without extra confirmation, while acceptance remains human-only. Memory capture policy: save/update only durable, future-relevant, actionable, scoped, evidence-backed, compact, non-sensitive observations; prefer update by known id/topicKey over duplicate saves; never save full SDD artifacts, transcripts, raw logs, hidden reasoning, secrets, credentials, or sensitive user data. Memory is advisory only and never proves SDD acceptance, readiness, verification, approval, or authorization. SDD artifacts are canonical for phase deliverables/governance state. Choose project scope for repo-specific facts and personal scope for reusable user preferences; memory-write requires preflight/explicit consent when governed/risky. Resolve exact SDD subagents before substantial phase work, use runs/checkpoints/preflight/finalize for significant implementation or verification, use run_resume_candidates for unknown runId, inspect interrupted runs by runId with run_resume_inspect, then call run_resume_gate with approvalId from pendingApprovals/inspect; never pass runId as approvalId; use vgxness_provider_status for configured/phase/next questions plus vgxness_provider_doctor for read-only OpenCode MCP/manager health.',
     'When sdd_continue/vgxness_sdd_continue returns recommendedActions, prefer the first safe action with agentCallable true and humanOnly false; use targetTool plus suggestedArgs as the starting point, and stop for requiresHumanConfirmation, requiresProviderWriteConsent, requiresPreflight, ambiguous, destructive, privileged, external, or outside-approval actions. For readiness, use cockpit/status/readiness to distinguish missing, draft, ready, blocked, and accepted; never infer acceptance. For provider evidence, report providerEvidence.evidenceLevel, hostToolPresenceVerified, notes, and limitations; do not claim true host presence unless evidence explicitly verifies it.',
     'Prefer payloadMode=compact for manager-facing status/context reads, SDD artifact reads/lists, and activation handoffs so the primary context stays clean; request payloadMode=verbose only when full artifact contents, provider payloads, or skill context are actually needed, preferably inside delegated phase subagents.',
     'MCP sdd_continue must not execute providers, create runs, mutate artifacts, write provider config/openspec, bypass human acceptance, or treat draft-run as apply-progress.',

package/docs/glossary.md

@@ -90,6 +90,8 @@ Model Context Protocol. The agent-facing transport. VGXNESS exposes 42 typed too
 
 A durable record in the SQLite store. Identified by `id`; upserted by `topicKey`. Types: `architecture`, `decision`, `bugfix`, `pattern`, `config`, `discovery`, `learning`, `preference`, `manual`. Project or personal scope.
 
+Memory observations should be durable, actionable, scoped, evidence-backed, compact, and non-sensitive. They are advisory only: memory never proves SDD acceptance, readiness, verification, approval, or authorization, and SDD artifacts are canonical for phase deliverables and governance state. Prefer update over duplicate save when a known `id` or `topicKey` exists. Do not save secrets, credentials, hidden reasoning, raw large logs, full transcripts, full SDD artifacts, or sensitive user data.
+
 ## Natural-language planner
 
 The provider-agnostic front-door classifier for operator text. Maps an intent to exactly one preview flow: `direct`, `plan`, `sdd`, or `diagnose`. Non-executing by design.

package/docs/mcp.md

@@ -39,6 +39,16 @@ All tools:
 
 `topicKey` is the durable upsert key: re-saving with the same key updates the existing observation rather than creating a duplicate. For `vgxness_memory_search` only, `topicKey: "*"` searches across all topic keys, and a top-level key such as `"release"` also matches observations under that prefix, for example `release/...`.
 
+### Durable memory capture policy
+
+Use memory for durable, future-relevant facts that are reusable beyond the current turn. A good memory is durable, actionable, scoped, evidence-backed, compact, and non-sensitive. Prefer update over duplicate save when a known `id` or `topicKey` already exists.
+
+- **Good:** project-scoped `decision` — "Use Bun for installed CLI/MCP runtime; Node remains development tooling" with a source file or docs reference.
+- **Good:** personal-scoped `preference` — "Prefer concise SDD apply summaries with changed files and verification evidence."
+- **Bad:** full phase artifacts, full transcripts, raw large logs, hidden reasoning, secrets, credentials, or sensitive user data.
+
+Memory is advisory only: it never proves SDD acceptance, readiness, verification, approval, or authorization. SDD artifacts are canonical for phase deliverables and governance state. Active-work memory remains advisory and should point to current context without replacing artifacts or runs. `memory-write` is a durable governed action; use preflight or explicit consent whenever policy marks it governed/risky.
+
 ## Sessions (5)
 
 | Tool | Purpose | Required inputs | Notes |

package/docs/storage.md

@@ -77,6 +77,14 @@ The schema follows a small set of conventions:
 - **Skills** are versioned, with `currentVersionId` pointing at the active version. Skill improvement proposals do not mutate the active skill until `applySkillImprovementProposal(...)` is called on an `approved` proposal.
 - **Manager profile overlays** are a thin layer over the canonical agent manifest; they record per-user/per-project model and prompt-contract preferences without mutating the built-in manifest.
 
+## Durable memory capture boundary
+
+Memory observations are for reusable context, not raw work logs. Capture only information that is durable, future-relevant, actionable, scoped, evidence-backed, compact, and non-sensitive. Use project-scoped memory for repo-specific decisions, discoveries, bug fixes, config, and patterns. Use personal-scoped memory for cross-project preferences and reusable user-level context.
+
+Prefer update over duplicate save: if a known observation `id` or `topicKey` already represents the fact, update that record so revision history stays attached to one durable topic. Do not save secrets, credentials, hidden reasoning, raw large logs, full transcripts, full SDD artifacts, or sensitive user data.
+
+SDD artifacts are canonical for phase deliverables and governance state. Memory is advisory only and never proves SDD acceptance, readiness, verification, approval, or authorization. Active-work memory remains advisory; it can summarize current context, but it does not replace runs, checkpoints, or artifacts. `memory-write` is a durable governed action and requires preflight/explicit consent when policy marks it governed or risky.
+
 ## Backup and recovery
 
 There is no automatic backup of the SQLite database. The recommended pattern is to treat the database file as a normal file in the user's data directory and use OS-level snapshots if needed.

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "vgxness",
-  "version": "1.19.3",
+  "version": "1.19.4",
   "description": "CLI and MCP control plane for guided AI-agent workflows, SDD, memory, and OpenCode setup.",
   "license": "SEE LICENSE IN LICENSE",
   "repository": {