Add easy-memory skill

[memeda230606]

Summary

• simplify the experimental easy-memory memory-agent contract so it only filters out unrelated memory blocks and appends a final [SUMMARY] line
• replace the older strict JSON response path with lightweight plain-text handling across the canonical prompt, response/output references, runtime client, and read/search scripts
• keep provider flexibility intact for codex_exec, OpenAI-compatible chat completions, and Ollama native chat, while preserving raw fallback plus installation-directory failure logging on agent-side errors
• update the request/output examples and smoke-test behavior so the canonical success path now expects retained original memory blocks followed by a final summary line
• preserve backward compatibility for existing related-resource naming while synchronizing the updated skill behavior to the global Codex installation

Tests

• python3 -m py_compile skills/.experimental/easy-memory/scripts/*.py
• python3 skills/.experimental/easy-memory/scripts/smoke_test_memory_agent.py --strict-no-new-failures --task-context "Verify the plain-text easy-memory agent redesign before commit and push." --search-keyword plain-text --search-keyword summary --search-keyword memory-agent

[memeda230606]

@gverma-openai Could you please take a look and review this PR? Thanks!

[github-actions]

All contributors have signed the CLA ✍️ ✅
_{Posted by the CLA Assistant Lite bot.}

[memeda230606]

I have read the CLA Document and I hereby sign the CLA

[memeda230606]

Updated this PR with the latest easy-memory agent integration work and pushed commit c354b42.

Summary of the latest changes:

• added optional memory-agent preprocessing with strict raw fallback on any agent-side error
• added canonical agents/, references/, and assets/examples/ content for the upstream-compatible package
• added related resource support for absolute local paths and URL/document addresses, including stable resource IDs and update flows
• added memory-agent config loading, OpenAI-compatible and Ollama native client support, installation-directory failure logging, and a smoke-test script
• fixed .gitignore so local runtime data stays ignored while the experimental skill source stays versioned

Tests run:

• python3 -m py_compile skills/.experimental/easy-memory/scripts/*.py
• python3 skills/.experimental/easy-memory/scripts/smoke_test_memory_agent.py --strict-no-new-failures --task-context "Smoke test for the current easy-memory memory-agent configuration in this project."

@gverma-openai @Slaymish Could you please take another look when you have a chance? Thank you.

[memeda230606]

Pushed commit baa6201 with the Codex CLI exec provider update for experimental easy-memory.

Highlights:

• added a codex_exec provider that runs codex exec locally with strict schema-constrained JSON output
• added configurable Codex model/service-tier/reasoning/profile settings, with defaults centered on gpt-5.3-codex-spark
• kept OpenAI-compatible and Ollama-native providers intact, while extending failure logging and smoke-test reporting with provider-specific config details
• updated the Codex-facing docs and example local config so Codex environments can prefer codex_exec

Validation run before push:

• python3 -m py_compile skills/.experimental/easy-memory/scripts/*.py
• python3 skills/.experimental/easy-memory/scripts/smoke_test_memory_agent.py --strict-no-new-failures --task-context "Smoke test for the codex_exec easy-memory configuration before commit and push."

Latest local smoke-test results:

• search_memory.py: 8.28s, status ok
• read_today_log.py: 7.21s, status ok
• failure log growth: 6 -> 6

PR status check after push:

• head SHA: baa6201
• mergeable state: blocked
• check runs reported by GitHub on the PR head: 0

@gverma-openai @Slaymish Please take a look when you have a chance. Thank you.

[memeda230606]

Pushed commit 100819f with the latest easy-memory update.

Highlights:

• simplified the memory-agent contract from strict JSON output to plain-text filtering with a final [SUMMARY] line
• updated the canonical prompt, response/output references, request/output examples, runtime client, read/search scripts, and smoke test to match the new flow
• kept agent-side fallback and installation-directory failure logging intact while preserving provider support for codex_exec, OpenAI-compatible HTTP, and Ollama native chat

Validation run before push:

• python3 -m py_compile skills/.experimental/easy-memory/scripts/*.py
• python3 skills/.experimental/easy-memory/scripts/smoke_test_memory_agent.py --strict-no-new-failures --task-context "Verify the plain-text easy-memory agent redesign before commit and push." --search-keyword plain-text --search-keyword summary --search-keyword memory-agent

Latest local smoke-test results:

• search_memory.py: 12.23s, status ok
• read_today_log.py: 9.09s, status ok
• failure log growth: 17 -> 17

Current PR status after push:

• head SHA: 100819f
• mergeable state: blocked
• last recorded approval is still tied to older commit 165d701

@gverma-openai @Slaymish Please take another look when you have a chance. Thank you.

[memeda230606]

I checked the current PR state after the latest push.

Current signals suggest two likely blockers:

• the only recorded approval is still tied to the older commit 165d701, not the current head 100819f
• after fetching upstream/main on 2026-03-18, this branch is now 13 ahead / 49 behind relative to the latest upstream main tip

The latest head already contains the plain-text easy-memory agent update and local validation is green, but if this repository requires an up-to-date branch and a fresh approval on the current head, that would explain the current mergeable_state = blocked.

@gverma-openai @Slaymish Please let me know if you would prefer this branch to be rebased or merged with the latest main before the next review. Thank you.

[memeda230606]

Synced pr/skills-update with the latest upstream/main and pushed merge commit 895603e.

Current branch comparison after the merge:

• pr/skills-update...upstream/main = 14 ahead / 0 behind
• PR head SHA = 895603e74d500eb9f318a20ad28895623ebf1990

Validation after the merge:

• python3 -m py_compile skills/.experimental/easy-memory/scripts/*.py
• python3 skills/.experimental/easy-memory/scripts/smoke_test_memory_agent.py --strict-no-new-failures --task-context "Verify the plain-text easy-memory agent redesign after merging upstream/main." --search-keyword plain-text --search-keyword summary --search-keyword memory-agent

Latest post-merge smoke-test results:

• search_memory.py: 9.32s, status ok
• read_today_log.py: 11.15s, status ok
• experimental failure log growth: 6 -> 6

The branch is now up to date with main, but GitHub still reports mergeable_state = blocked. At this point the remaining likely blocker is that the only recorded approval still targets older commit 165d701, not the current head.

@gverma-openai @Slaymish Please take another look when you have a chance. Thank you.