skills/codex/memex-search/SKILL.md
Search, filter, and retrieve Claude/Codex history indexed by the memex CLI. Use when you want to search history, run lexical/semantic/hybrid search, fetch full transcripts, or produce LLM-friendly JSON output.
npx skillsauth add nicosuave/memex memex-searchInstall this skill globally with one command. Works with Claude Code, Cursor, and Windsurf.
3 of 9 scanners reported clean
Some scanners were skipped, did not run, or reported a non-clean status. Review each row below.
Use this skill to index local history and retrieve results in a structured way.
memex indexmemex index-service enable --continuousmemex reindexmemex index --no-embeddingsmemex embed--source <path> for Claude logs--include-agents to include agent transcripts--codex/--no-codex to include or skip Codex logs--model <minilm|bge|nomic|gemma|potion> to select embedding model--root <path> to change data root (default: ~/.memex)Run a search; output is JSON lines by default.
memex search "query" --limit 20
Each JSON line includes:
doc_id, ts (ISO), session_id, project, role, source_pathtext (full record text)snippet (trimmed single-line summary)matches (offsets + before/after context)score (ranked score)| Need | Command |
| --- | --- |
| Exact terms | search "exact term" |
| Fuzzy concepts | search "concept" --semantic |
| Mixed | search "term concept" --hybrid |
--project <name>--role <user|assistant|tool_use|tool_result>--tool <tool_name>--session <session_id> (search inside a transcript)--source claude|codex--since <iso|unix> / --until <iso|unix>--limit <n>--min-score <float>--top-n-per-session <n> (top n per session)--unique-session (same as top-k per session = 1)--sort score|ts (default score)--json-array for a single JSON array--fields score,ts,doc_id,session_id,snippet to reduce output-v/--verbose for human outputmemex index-service enable
memex index-service enable --continuous
memex index-service disable
--limit--project and --since/--until--top-n-per-session or --unique-sessionmemex session <id> for full context--unique-session to collapse PR-link spam fast.--min-score to prune low-signal hits.--sort ts when you want a timeline view.--role assistant for narrative outcomes; --role tool_result for command errors.search "<term>" --session <id> --sort ts --limit 50 to jump to outcomes.Create ~/.memex/config.toml (or <root>/config.toml if you use --root):
embeddings = true
auto_index_on_search = true
model = "potion" # minilm, bge, nomic, gemma, potion
scan_cache_ttl = 3600 # seconds (default 1 hour)
index_service_mode = "interval" # interval or continuous
index_service_interval = 3600 # seconds (ignored when mode = "continuous")
index_service_poll_interval = 30 # seconds
auto_index_on_search runs an incremental index update before each search.
scan_cache_ttl sets the maximum scan staleness for auto-indexing.
index-service reads config defaults (mode, interval, log paths). Flags override.
Service logs and the plist live under ~/.memex by default.
Recommended when embeddings are on (especially non-potion models): run the
background index service or index --watch, and consider setting
auto_index_on_search = false to keep searches fast.
--semantic--hybrid--recency-weight <float>--recency-half-life-days <float>memex show <doc_id>memex session <session_id>Both commands return JSON by default.
Use -v/--verbose for human-readable output:
memex search "query" -vmemex show <doc_id> -vmemex session <session_id> -vmemex search "query" --limit 20matches or snippetmemex show <doc_id> or memex session <session_id>--session, --role, or time filtersdevelopment
Search memex for user feedback patterns (frustration, corrections, praise, successful outcomes) to identify recurring mistakes and wins, then generate CLAUDE.md or AGENTS.md improvements.
tools
Use when work should span one or more detached tasks but still behave like one job with a single owner context. TaskFlow is the durable flow substrate under authoring layers like Lobster, ACPX, plugins, or plain code. Keep conditional logic in the caller; use TaskFlow for flow identity, child-task linkage, waiting state, revision-checked mutations, and user-facing emergence.
tools
# Lobster Lobster executes multi-step workflows with approval checkpoints. Use it when: - User wants a repeatable automation (triage, monitor, sync) - Actions need human approval before executing (send, post, delete) - Multiple tool calls should run as one deterministic operation ## When to use Lobster | User intent | Use Lobster? | | ------------------------------------------------------ | --------------------------
tools
# Lobster Lobster executes multi-step workflows with approval checkpoints. Use it when: - User wants a repeatable automation (triage, monitor, sync) - Actions need human approval before executing (send, post, delete) - Multiple tool calls should run as one deterministic operation ## When to use Lobster | User intent | Use Lobster? | | ------------------------------------------------------ | --------------------------