Memory Recall
Forked subagent that recalls past-session decisions via the memsearch semantic-memory CLI.
Test report
- Verdict
- Works with setup
- Score
- Tested
- Jul 20, 2026
- Environment
- Claude Code 2.x (agent harness)
- Upstream re-checked
- Aug 10, 2026 · 9f5e70d
Fetched the repo tree (a 4-plugin monorepo: claude-code/codex/openclaw/opencode) and read the claude-code variant's SKILL.md, plugin.json (v0.4.15), README, and derive-collection.sh raw. SKILL.md frontmatter is valid (name+description, context: fork, allowed-tools: Bash) and both referenced files (scripts/derive-collection.sh, transcript.py) exist in the tree; no exfiltration/base64/secrets found — the script just SHA-hashes the project path into a Milvus collection name. Docked 1 install point because the copied skill is inert without the external memsearch Python CLI (auto-installed via uvx on first run) plus a populated .memsearch store. Trigger 5/5 tested phrasings — SHOULD: (1) "What did I decide about the auth token refresh approach last week?" (2) "Have we hit this Milvus connection error before?" (3) "Why did we switch from JWT to session cookies?"; should NOT: (4) "What functions are defined in src/auth.py right now?" (current state -> Read/Grep) (5) "Fix the typo on this line I'm looking at" (ephemeral) — all 5 judged correctly. OUTPUT NOT measured (not a measured loss): requires an external CLI I won't install and cross-session memory data that doesn't exist here, so output=5 as unmeasured-neutral. Verdict setup: legitimate, well-documented, honest about the uvx/CLI dependency, but real friction (external memsearch CLI + background watch process + accumulated memory).
Scored on four weighted criteria — install, triggering, output vs. baseline, docs. How scoring works
- Installs cleanly 4/5
- Triggers reliably 5/5
- Output vs. baseline 5/10
- Docs & honesty 5/5
What Memory Recall does
A retrieval skill that runs as a forked subagent, searching a local semantic memory index of past Claude Code sessions to surface prior decisions, debugging notes, and project context. Triggers on questions like "what did I decide about X" or "why did we do Y", or on injected [memsearch] Memory available hints. Skips when the question is purely about current code state.
How to install Memory Recall
/plugin marketplace add zilliztech/memsearch && /plugin install memsearch (restart Claude Code; first run auto-installs the memsearch[onnx] CLI via uvx)
Skills live in ~/.claude/skills/ (global) or .claude/skills/
(per-project). Restart Claude Code after installing.
Commands — how to trigger Memory Recall
-
/memory-recallForked subagent that recalls past-session decisions via the memsearch semantic-memory CLI.
It also activates on plain-language prompts like these:
-
What did I decide about the database schema before -
Have I seen this error before in past sessions -
Why did we choose this approach last time
Frequently asked questions
- Is the Memory Recall skill free?
- Yes. The skill itself is free from zilliztech/memsearch. SkillProof publishes the install command and an independent test verdict at no cost.
- Does Memory Recall work with Claude Code?
- We tested it with Claude Code 2.x (agent harness) on Jul 20, 2026. Verdict: Works with setup. Fetched the repo tree (a 4-plugin monorepo: claude-code/codex/openclaw/opencode) and read the claude-code variant's SKILL.md, plugin.json (v0.4.15), README, and derive-collection.sh raw. SKILL.md frontmatter is valid (name+description, context: fork, allowed-tools: Bash) and both referenced files (scripts/derive-collection.sh, transcript.py) exist in the tree; no exfiltration/base64/secrets found — the script just SHA-hashes the project path into a Milvus collection name. Docked 1 install point because the copied skill is inert without the external memsearch Python CLI (auto-installed via uvx on first run) plus a populated .memsearch store. Trigger 5/5 tested phrasings — SHOULD: (1) "What did I decide about the auth token refresh approach last week?" (2) "Have we hit this Milvus connection error before?" (3) "Why did we switch from JWT to session cookies?"; should NOT: (4) "What functions are defined in src/auth.py right now?" (current state -> Read/Grep) (5) "Fix the typo on this line I'm looking at" (ephemeral) — all 5 judged correctly. OUTPUT NOT measured (not a measured loss): requires an external CLI I won't install and cross-session memory data that doesn't exist here, so output=5 as unmeasured-neutral. Verdict setup: legitimate, well-documented, honest about the uvx/CLI dependency, but real friction (external memsearch CLI + background watch process + accumulated memory).
- What is the Memory Recall SkillProof Score?
- 7.6/10 — installs cleanly 4/5, triggers reliably 5/5, output vs. baseline 5/10, docs & honesty 5/5.
- How do I install Memory Recall?
- Copy the install command from this page, run it in your terminal, and restart Claude Code. Skills live in ~/.claude/skills/ (global) or .claude/skills/ inside a project.
- Can I use Memory Recall with Cursor, Copilot, Gemini CLI, Codex or other AI tools?
- The SKILL.md format is native to Claude (Claude Code, Desktop, claude.ai). The instructions inside adapt to other assistants: Cursor rules, GitHub Copilot instructions, Windsurf rules, Custom GPTs, AGENTS.md for OpenAI Codex, and GEMINI.md for Google Gemini CLI — our conversion guides cover each, and the free converter on the tools page does the wrapping for you.