skills-generator Memory Assistant
Workflow
- Record key context immediately after meaningful events (decisions, fixes, assumptions).
- Tag entries for retrieval (
bugfix,decision,todo,risk,context). - Retrieve memory by keyword/tag before continuing complex tasks.
- Generate short recall digests to bootstrap future sessions quickly.
Commands
Add memory
python3 scripts/add_memory.py \
--file ./references/memory.md \
--title "SAT threshold changed" \
--note "Raised threading threshold to K>=1200 after regression in medium cases." \
--tags "decision,performance" \
--source "benchmark-2026-04-15"
Search memory
python3 scripts/search_memory.py "threshold|thread" --file ./references/memory.md --ignore-case
Recall digest
python3 scripts/recall_digest.py --file ./references/memory.md --count 8 --out ./references/recall-digest.md
Entry Quality Rules
- Keep title short and specific.
- Keep note factual and actionable.
- Prefer one decision per entry.
- Always add tags when possible.
- Mark uncertainty explicitly in note text.
Retrieval Rules
Before making major changes:
- Search memory for prior decisions and constraints.
- Summarize conflicting history before selecting approach.
- If memory is stale, append a fresh entry documenting supersession.
Resource Usage
- Use memory-template.md for consistent entry style.
- Use add_memory.py to persist memory.
- Use search_memory.py for targeted recall.
- Use recall_digest.py for quick session bootstrap.
Safety Notes
- Do not store secrets, tokens, credentials, or private personal data.
- Keep entries project-relevant and least-privilege in detail.