Psychology Plugin v1.2.21

Psychology self-reflection plugin package and host-configuration contract.\n\nVerified: npm test passes the public package contract (62 public files, 14 root skills, 43 negative fixtures). This release does not assert marketplace approval, authenticated host acceptance, or account-history retrieval. The hosted psychology MCP service must be independently available and authenticated.

Psychology Plugin v1.2.20

Memory-first source package: distill only the memories and past chats the current ChatGPT or Claude session can actually access; Codex remains limited to this task and user-selected or pasted material. The prompt and manifest checks preserve evidence, temporal uncertainty, consent and no-save boundaries. The live catalog at release time contains 34 psychology-prefixed tools and 18 prompts. This is a GitHub source release only; it does not establish full archive access or official ChatGPT/Claude marketplace acceptance.

Psychology Plugin v1.2.19

Correct the documented canonical OAuth protected resource and add an offline verifier assertion so the public README and connector manifests stay aligned at https://noesis.seges.ai/mcp. Bump all plugin package manifests to 1.2.19. This source release changes documentation and validation only; it does not change or verify live OAuth behavior.

Psychology Plugin 1.2.18

Consent-safe assessment and journal guidance. The assessment guide now requires an authenticated tool-list check and user-requested eligible self-report before any versioned result save. MCP cannot create or expand journal grants; users manage those permissions in the signed-in Noesis website. The contract verifier prevents retired write-tool guidance from returning.

Psychology Plugin v1.2.17

Quick-start distillation now asks at most three follow-up questions, and only when each answer could change the next step. The canonical prompt and Codex mirror verifier enforce the rule. This release does not imply host approval or broader historical-chat access.

Psychology Plugin v1.2.16

Caps ChatGPT, Claude, and Kimi selected-memory summary chains at 32 parts (3,200 items), matching the Noesis import contract. The prompt preserves incomplete status at the cap and asks users to narrow the source selection; the verifier rejects longer chains. This source release does not claim full archive access, marketplace acceptance, or additional host support.

Psychology Plugin v1.2.15

Source release: strengthens the memory-first quick-start with distinct event, memory-save, summary-generation, claim-applicability, and Noesis-receipt clocks. Current corrections update present status without inventing when an older goal ended. Preserves ChatGPT, Claude, and Kimi selected-summary v2; Codex remains current-task-only v3. Verification: npm test passed. This is a source release only; it does not claim marketplace acceptance or expanded host support.

Psychology Plugin v1.2.14

Source release. Adds traceable source notes to memory-distillation candidates and separates source event time, memory update time, summary generation time, applicability, and Noesis receipt time when available. Marks unavailable references and dates unknown rather than inventing them. Contract verification passed locally (90 public files, 28 mirrored skill files). This release does not claim ChatGPT or Claude marketplace acceptance.

Psychology Plugin v1.2.13

Clarifies that cross-assistant context uses a purpose-limited selected-memory summary, not a full account archive export. Caps guide-failure fallback follow-up questions at three consequential gaps. Portable plugin contract tests pass. Source release only; this does not claim ChatGPT or Claude marketplace acceptance.

Psychology Plugin v1.2.12

Adds an explicit, owner-approved selected-context import flow. Ordinary distillation remains limited to context available in the current assistant conversation. Imports require current consent, listing the current connection's active scoped grants, the user choosing one returned grant, and reading only within its exact purpose, categories, date range, and time basis. Missing or invalid grants fail closed.

This GitHub source release does not claim official ChatGPT or Claude marketplace acceptance.