How do I install the Agent Memory Ledger skill?

Run npx skillmds@latest add yufanyang214-lab/agent-memory-ledger in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

Is Agent Memory Ledger safe to use?

This skill has not completed SkillMD's automated safety review yet. Capability flags: executes scripts. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with Agent Memory Ledger?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is Agent Memory Ledger free to use?

Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.

Who published Agent Memory Ledger?

yufanyang214-lab (@yufanyang214-lab) published this skill. Their other Agent Skills are listed on their SkillMD profile.

Agent Memory Ledger by yufanyang214-lab

Name: Agent Memory Ledger
Rating: 4.2 (5 reviews)
Author: yufanyang214-lab

File contents

Agent Memory Ledger for Codex CLI

Use this skill when the user asks Codex to remember, archive, or recall durable information across separate Codex sessions.

Boundary

The standalone ledger is the canonical memory write target. Do not enable or call Codex experimental memories, and do not copy memory bodies into AGENTS.md or other startup instructions. An external memory provider may receive only sanitized evidence or derived objects after successful AML archival when the operator explicitly configured external-memory mode. The adapter reads only visible user and assistant events from the current rollout. It excludes developer instructions, hidden reasoning, tool calls, command output, plugin recommendations, and model metadata.

Resolve SKILL_DIR as the absolute directory containing this loaded SKILL.md. Codex exposes the skill location in its skill metadata; do not assume that the skill came from .agents/skills, because a plugin installation uses the Codex plugin cache.

Install the core package and stable repository-local bridge once:

python "$SKILL_DIR/scripts/install.py" \
  --workspace "$PWD" \
  --source /path/to/agent_memory_ledger-0.1.1-py3-none-any.whl

Use these paths from the repository root:

PY="$PWD/.agent-memory-ledger-venv/bin/python"
BRIDGE="$PWD/.agent-memory-ledger-adapter/codex_bridge.py"
LEDGER="$PWD/.portable-memory"

Archive the current Codex session

CODEX_THREAD_ID is provided to shell commands by Codex CLI, so the bridge can locate the current rollout without asking the user for a session path.

Archive only when the user affirmatively asks to archive the current session with the configured trigger or an equivalent such as "archive", "archive this session", “归档本会话”, or “按 AML 归档”. Questions, explanations, quotations, hypotheticals, and negations about archiving do not trigger the workflow. Before running the bridge, extract durable knowledge, procedure, and event candidates into a JSON file as described in docs/AGENT_INSTRUCTIONS.md. Use tags for finer subtypes and canonical kind names only. Set promote: true only for the small set of stable, high-value items that should be discoverable at startup.

"$PY" "$BRIDGE" archive \
  --ledger "$LEDGER" \
  --candidate-file /path/to/memory-candidates.json

Every accepted active candidate is represented in the dark ledger. The bridge uses candidate promote values to build the bright ledger. Add --promote-markers only when the user explicitly supplied stable lines such as Fact:, Rule:, Procedure:, Decision:, 事实：, 规则：, or 流程：. Never edit evidence, objects, ledgers, SQLite, or the journal directly. Run validation after archival and report the archive, object, and promoted IDs.

Recall

"$PY" "$BRIDGE" recall \
  --ledger "$LEDGER" \
  --query "<user question>" \
  --top-k 6

Answer only from returned active objects and include relevant object_id values. Do not silently fall back to session history, Codex memories, or startup files. An explicitly configured external provider may be used for deeper recall, but its result must retain or resolve to AML archive/object provenance.

Validate

"$PY" "$BRIDGE" validate --ledger "$LEDGER"

Completion requires ok: true, evidence, structured objects, both ledgers, and successful recall from a fresh Codex thread.

---name: agent-memory-ledgerdescription: Archive or recall durable facts, procedures, decisions, and events across Codex CLI sessions using a standalone evidence/object/bright-dark-ledger workspace rather than Codex memories.---# Agent Memory Ledger for Codex CLIUse this skill when the user asks Codex to remember, archive, or recall durableinformation across separate Codex sessions.## BoundaryThe standalone ledger is the canonical memory write target. Do not enable orcall Codex experimental memories, and do not copy memory bodies into `AGENTS.md`or other startup instructions. An external memory provider may receive onlysanitized evidence or derived objects after successful AML archival when theoperator explicitly configured external-memory mode. The adapter reads onlyvisible user and assistant events from the current rollout. It excludesdeveloper instructions, hidden reasoning, tool calls, command output, pluginrecommendations, and model metadata.Resolve `SKILL_DIR` as the absolute directory containing this loaded`SKILL.md`. Codex exposes the skill location in its skill metadata; do notassume that the skill came from `.agents/skills`, because a plugin installationuses the Codex plugin cache.Install the core package and stable repository-local bridge once:```bashpython "$SKILL_DIR/scripts/install.py" \ --workspace "$PWD" \ --source /path/to/agent_memory_ledger-0.1.1-py3-none-any.whl```Use these paths from the repository root:```bashPY="$PWD/.agent-memory-ledger-venv/bin/python"BRIDGE="$PWD/.agent-memory-ledger-adapter/codex_bridge.py"LEDGER="$PWD/.portable-memory"```## Archive the current Codex session`CODEX_THREAD_ID` is provided to shell commands by Codex CLI, so the bridge canlocate the current rollout without asking the user for a session path.Archive only when the user affirmatively asks to archive the current sessionwith the configured trigger or an equivalent such as "archive", "archive thissession", “归档本会话”, or “按 AML 归档”. Questions, explanations, quotations,hypotheticals, and negations about archiving do not trigger the workflow.Before running the bridge, extract durable `knowledge`, `procedure`, and`event` candidates into a JSON file as described in`docs/AGENT_INSTRUCTIONS.md`. Use tags for finer subtypes and canonical kindnames only. Set `promote: true` only for the small set of stable, high-valueitems that should be discoverable at startup.```bash"$PY" "$BRIDGE" archive \ --ledger "$LEDGER" \ --candidate-file /path/to/memory-candidates.json```Every accepted active candidate is represented in the dark ledger. The bridgeuses candidate `promote` values to build the bright ledger. Add`--promote-markers` only when the user explicitly supplied stable lines such as`Fact:`, `Rule:`, `Procedure:`, `Decision:`, `事实：`, `规则：`, or `流程：`.Never edit evidence, objects, ledgers, SQLite, or the journal directly. Runvalidation after archival and report the archive, object, and promoted IDs.## Recall```bash"$PY" "$BRIDGE" recall \ --ledger "$LEDGER" \ --query "<user question>" \ --top-k 6```Answer only from returned active objects and include relevant `object_id` values.Do not silently fall back to session history, Codex memories, or startup files.An explicitly configured external provider may be used for deeper recall, butits result must retain or resolve to AML archive/object provenance.## Validate```bash"$PY" "$BRIDGE" validate --ledger "$LEDGER"```Completion requires `ok: true`, evidence, structured objects, both ledgers, andsuccessful recall from a fresh Codex thread.

Frequently asked questions

Agent Memory Ledger