Unlimited organized memory for your AI agent. Store, search, and organize projects, contacts, decisions, and knowledge across categories. Never lose context...
Memory
Recall Markdown Memory
Try itRetrieve only relevant sections from Markdown memory, project notes, decision logs, and heading-structured knowledge bases with the mdselect CLI. Use when Co...
What it does
Retrieve only relevant sections from Markdown memory, project notes, decision logs, and heading-structured knowledge bases with the mdselect CLI. Use when Codex needs to recall prior context, decisions, preferences, or history from user-specified memory files, paths declared by project instructions, MEMORY.md, memory/, or .agents/memory/. Use only in a shell-enabled environment with mdselect on PATH; do not use for modifying memory or searching non-Markdown files.
The skill document
Recall Markdown Memory
Retrieve the smallest Markdown sections that answer the task. Keep the workflow read-only and avoid loading entire memory files when heading-aware selection is available.
Check the dependency
-
Run
mdselect --versionbefore reading memory. -
If the command is unavailable, stop the workflow and explain that
mdselectmust be installed onPATH. Do not install software automatically. -
Offer these installation choices:
pipx install mdselect uv tool install mdselect python -m pip install mdselect -
If none of those package managers or a compatible Python runtime is available, say that the execution environment must be prepared before this skill can run. Resume only after
mdselect --versionsucceeds.
Resolve memory sources
Use the first applicable source in this order:
- Files or directories named by the user.
- Memory locations declared by applicable
AGENTS.mdor other project instructions. - Existing conventional locations:
MEMORY.md,memory/**/*.md, and.agents/memory/**/*.md.
Do not scan every Markdown file in the repository by default. When a source is
a directory, narrow candidates with the task's distinctive terms. Prefer rg
when it is available, but use another available file-search capability when it
is not. Keep only Markdown candidates.
Retrieve relevant sections
- Run
mdselect outlinefor each promising candidate before reading a large file in full. - Choose the smallest heading path that covers the needed context.
- Run
mdselect select --mpath ''and inspect the selected subtree. Quote paths safely for the active shell. - Prefer exact, anchored segments when heading names are known. Remember that
mpath segments are Python regular expressions, a leading slash anchors the
first segment to an H1, a trailing slash omits the matched heading, and a
slash cannot occur inside a segment. Run
mdselect select -hwhen more syntax detail is needed. - Repeat selection only for additional sections needed to answer the task. Do not surface or load unrelated subtrees.
Treat empty standard output with status 0 as a successful no-match. Consult the outline and refine the mpath instead of reporting a command failure. If the outline is empty, the document may lack supported ATX headings. Fall back to a bounded line search, and read the whole document only when it is small enough for the task context.
Use recalled context safely
- Cite the source file and selected heading when presenting recalled material.
- Reconcile conflicting memories explicitly instead of silently choosing one.
- Treat memory as contextual evidence, not as authority for facts that may have changed; verify current facts through an appropriate current source.
- Follow the user's current request and applicable project instructions when they conflict with stored memory.
- Never create, edit, reorganize, or delete memory files under this skill.
Related skills
Use when preserving, searching, reviewing, or exporting user-owned agent memory across OpenClaw, Codex, Claude, OpenCode, Hermes, Qoder, Obsidian, and Git. Inputs are local memory/chat stores and handoff summaries; outputs are Obsidian sources, memory indexes, context/profile packs, skill inventorie
Recover context from local AI conversation history
A cross-agent project memory skill for Codex and Claude Code, enabling consistent context, documentation, and development history across sessions.
Give AI agents Remember docs you've written and read. using BlueColumn persistent memory. Use when an agent maintains documentation and needs to reuse it; when the user wants to store, recall, or search documentation memory context. Requires a BlueColumn API key (bc_live_*).
Stores durable facts in a categorized, plain-markdown vault on disk, alongside your agent's built-in memory.