Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add commands/lijialex/layered-memory/buildgit clone --depth 1 https://github.com/LijiAlex/layered-memoryWhat it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00029 | $0.00447 |
| Opus 5 | $0.00015 | $0.00224 |
| Sonnet 5 | $0.00006 | $0.00089 |
| Haiku 4.5 | $0.00003 | $0.00045 |
Grade A, and why
build scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
What it actually says
Run the layered-memory build over your past session transcripts. This reads your
Claude Code transcripts (newest first, skipping ones already ingested), distils them
into per-theme summaries, and writes/updates ~/.claude/memory/themes/*.md and
~/.claude/memory/index.md. Existing summaries are reconciled (not overwritten blindly);
a snapshot is taken before each write.
Each transcript = one model call (~$0.05). Use --limit N to cap cost on the first
run, e.g. /memory:build --limit 3 ingests only the 3 newest new transcripts. Re-running
continues where it left off (a processed.log ledger skips done sessions).
Execute (forwards any arguments the user passed, e.g. --limit 3):
python3 "${CLAUDE_PLUGIN_ROOT}/scripts/build.py" $ARGUMENTS
[MUST] Run it in the FOREGROUND so the user sees progress. The script streams live
[memory] … milestone lines (scanning, per-transcript reading/distilling, theme counts,
reconcile). Therefore:
- Do NOT run it in the background.
- Do NOT pipe it to
tail,head, or anything that withholds output until completion. - Run it as-is and surface its streaming stdout to the user as it appears.
Note for the user: builds take ~10–40s per transcript (one model call each) plus a final
auto-reconcile call, so output appears gradually. For the most direct live view the user can
also run the command themselves with a leading ! in the prompt, or in a real terminal.
After it exits, report the themes written, the transcript count, any errors, and where the
index lives. Theme files are plain markdown under ~/.claude/memory/themes/ for inspection.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 2d ago First seen · 35 lines · 29 tokens per session scan A e62385c64597
build is a command published in the GitHub repository LijiAlex/layered-memory (1 stars, last pushed 2mo ago), licensed MIT. It adds 29 tokens to every session and 447 once invoked, about $0.0001 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other commands, from other repositories
git
Git operations with intelligent commit messages and workflow optimization.
checklist
Generate a custom checklist for the current feature based on user requirements.
clarify
Identify underspecified areas in the current feature spec by asking up to 5 highly targeted clarification questions and encoding answers back into the spec.
specify
Create or update the feature specification from a natural language feature description.
analyze
Perform a non-destructive cross-artifact consistency and quality analysis across spec.md, plan.md, and tasks.md after task generation.
converge
Assess the current codebase against the feature's spec, plan, and tasks, then append any remaining unbuilt work as new tasks to tasks.md so implement can complete it.