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/marcelopaniza/mempenny/memory-auto-splitgit clone --depth 1 https://github.com/marcelopaniza/mempennyWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/commands/marcelopaniza/mempenny/memory-auto-split)<a href="https://agentmods.dev/commands/marcelopaniza/mempenny/memory-auto-split"><img src="https://agentmods.dev/badge/commands/marcelopaniza/mempenny/memory-auto-split.svg" alt="Measured on agentmods" height="20"></a>What 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 | $0.00075 | $0.05197 |
| Opus 5 | $0.00037 | $0.02599 |
| Sonnet 5 | $0.00015 | $0.01039 |
| Haiku 4.5 | $0.00007 | $0.00520 |
Grade A, and why
memory-auto-split 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 3d 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.
How it starts
The opening of the file, as written. The whole thing — 152 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Split an over-ceiling topic file that has no other automated reduction path. This is the gap /mempenny:memory-curate and /mempenny:memory-shard-roll leave open (docs/memory-taxonomy-design.md §2-§4b): curate never touches charter.md/pending.md (no ### entries, and distilling requirements or in-flight work is destructive — correctly so), and shard-roll never touches a year that hasn't fully closed ("no mid-year shards, ever" is shard-roll's own, deliberate pin). Before this command existed, an over-ceiling pending.md (or a log-topic stuck entirely in the open year) had no automatic reduction path at all and just grew forever until a human noticed — see docs/memory-taxonomy-design.md's §4b for the incident that motivated this.
The unifying principle: over ceiling → the main file becomes an INDEX, and the items it used to hold become PAGES the index points at. Three strategies below are three expressions of that one rule, chosen mechanically from what the file actually contains — never guessed, never an LLM judgment call — plus one last-resort fallback for content with no structure to index at all:
- DATE (worklog/support/decisions) — the chronological expression. The parent's own
## Shardsindex block (Step 6, below) IS the index; the locked per-period shard files ARE the pages. - SUBJECT-INDEX (charter/pending, when
##-structured — the primary, real shape of a workingpending.md) — the subject expression, and the most direct one: the live file itself becomes the index outright (frontmatter + preamble + one bullet per subject), and every top-level##block becomes its own detail page. - PROSE-PEEL (charter/pending, only when neither of the above applies) — a last-resort positional split for genuinely structureless prose that still carries datestamps, no index/pages framing, just oldest-vs-newest.
- Refuse closed — nothing to safely index, peel, or order; reported for manual trimming.
Unlike curate, this command makes no judgment call about content — it never decides what's worth keeping. It is pure, mechanical, content-preserving relocation. That is what makes it safe to run unattended, even on pending.md: nothing is distilled, archived-by-verdict, or deleted — every byte that leaves the live file is proven to still exist, verbatim, somewhere in the pages.
Step 1 — Parse arguments
The user invoked this command with: $ARGUMENTS
- First positional argument — absolute path to a topic file. Required.
--lang <code>— output language. Default fromMEMPENNY_LOCALEoren.
There is no --yes flag and no confirmation gate — see "Why no confirmation gate" under Constraints below; this matches /mempenny:memory-shard-roll, not /mempenny:memory-curate.
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.
- 3d ago First seen · 152 lines · 75 tokens per session scan A 185fc191e918
memory-auto-split is a command published in the GitHub repository marcelopaniza/mempenny (2 stars, last pushed 14d ago), licensed MIT. It adds 75 tokens to every session and 5,197 once invoked, about $0.0004 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
mempalace-init
Set up MemPalace — install the package, initialize a palace, register the MCP server with Cursor, and verify everything works.
mempalace-status
Show the current state of your memory palace — wings, rooms, drawer counts, and suggestions.
ingest
Ingest source material into an active wiki. Accepts URLs, file paths, PDFs, freeform text, or processes the inbox. Supports tweets via Grok MCP.
fire-debug
Systematic debugging with persistent state, skills library integration, and WARRIOR validation.
review
Review the current diff against project memory.
speckit.archive
Archive a feature specification into main project memory after merge, resolving gaps and conflicts.