Append, rebuild, and verify records into a memory log for long-term project knowledge retention.

Install

mkdir -p .claude/skills/memory-write && curl -L -o skill.zip "https://agentskills.codes/api/skills/download/17936" && unzip -o skill.zip -d .claude/skills/memory-write && rm skill.zip

Installs to .claude/skills/memory-write

Activation

This is the description your AI agent reads to decide when to run this skill — the better it matches your request, the more reliably it fires.

Append a durable record (source or concept) to the governed memory log, then rebuild and verify.
96 charsno explicit “when” trigger
Beginner

Key capabilities

  • Draft a one-paragraph summary for the record
  • Collect source IDs for concept records
  • Write source or concept records to the memory log
  • Rebuild the governed memory index
  • Verify the integrity of the memory log

How it works

This skill appends durable records, either sources or concepts, to a governed memory log, requiring a summary and metadata, followed by a rebuild and verification of the index.

Inputs & outputs

You give it
A source worth keeping or a durable idea derived from sources
You get back
A durable record appended to the governed memory log, with a rebuilt and verified index

When to use memory-write

  • Persist a source document for future reference
  • Draft a concept record linking multiple source IDs
  • Categorize knowledge with tags like retrieval or design
  • Verify the integrity of the memory log after appending
  • Update the index with durable project insights

About this skill

Skill: memory-write

Use this skill when you need to persist something durable — a source you ingested, or a concept you derived. Writing is append-only and always followed by a rebuild and a verify.

When to use

  • You read a source worth keeping → write a source.
  • A durable idea emerged across sources → write a concept that cites them.

When not to use

  • For scratch thinking that does not need to survive the session. Not everything is memory. Over-writing pollutes the index with noise.

Steps

  1. Draft a one-paragraph summary. This is the navigation layer — make it stand on its own. If a reader cannot tell from the summary whether to open the body, rewrite it.

  2. For a concept, collect the ids of the sources it relies on.

  3. Write the record:

    governed-memory write \
        --type concept \
        --title "Summary-first navigation scales sublinearly" \
        --summary "Indexing summaries instead of bodies lets a reader decide what to open before paying the token cost." \
        --source <source-id> \
        --tag retrieval --tag design \
        --sensitivity public
    
  4. Rebuild and verify:

    governed-memory rebuild
    governed-memory verify     # must print OK before you stop
    

Guardrails

  • A source record must not declare --source.
  • --sensitivity restricted requires --acknowledge-restricted and explicit user consent.
  • If verify fails, fix the log (not the index) and rebuild.

When not to use it

  • For scratch thinking that does not need to survive the session
  • When over-writing the index with noise is a concern

Limitations

  • A `source` record must not declare `--source`
  • `--sensitivity restricted` requires explicit user consent
  • If `verify` fails, the log must be fixed, not the index

How it compares

This skill enforces a structured, append-only process for persisting knowledge with mandatory verification, ensuring integrity and discoverability, unlike informal note-taking.

Compared to similar skills

memory-write side by side with the closest alternatives in the catalog.

SkillInstallsUpdatedSafetyDifficulty
memory-write (this skill)01moNo flagsBeginner
braindump75moNo flagsBeginner
super-search101moReviewBeginner
mem-search83moNo flagsIntermediate

Try saying

Example prompts that trigger this skill in your AI assistant.

Search skills

Search the agent skills registry