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 skills/u4csolutions/midmem/midmem-ingestnpx skills add U4CSolutions/MidMem --skill midmem-ingestgit clone --depth 1 https://github.com/U4CSolutions/MidMemWrote 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/skills/u4csolutions/midmem/midmem-ingest)<a href="https://agentmods.dev/skills/u4csolutions/midmem/midmem-ingest"><img src="https://agentmods.dev/badge/skills/u4csolutions/midmem/midmem-ingest.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 | $0.00076 | $0.01152 |
| Opus 5 | $0.00038 | $0.00576 |
| Sonnet 5 | $0.00015 | $0.00230 |
| Haiku 4.5 | $0.00008 | $0.00115 |
Grade A, and why
midmem-ingest 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 4d 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 — 69 lines — stays where its author put it; the contents beside it link to each section on GitHub.
MidMem Ingest — verified knowledge intake
Commit a source into the store the RIGHT way: staged into an allowed root, grounding-checked (deterministic — DELEGATE-52 discipline), recall-verified, recorded. Never blind-ingest; never trust an extractor's own claim of faithfulness — read the grounding numbers.
Inputs this handles
- A file path already under one of your configured ingest source roots.
- A URL (arXiv page/PDF, blog post, article) → fetch it FIRST into a staging directory inside an allowed root (one subdir per source, kebab slug). For arXiv prefer the abs page or HTML render; keep the paper id in the filename.
- A recurring research digest (e.g. a weekly report of new papers on your domain). Ingest the
REPORT as one document (
--type research), then — only for items the report marks high-relevance or the operator names — fetch and ingest those sources individually.
Citations review (REQUIRED for reports/digests — before ingesting)
Recurring research digests carry citations; review the structure and build a manifest FIRST:
- Extract every citation link (
[Title](url)markdown). Canonicalize each URL: strip tracking params (?utm_source=…and friends); for arXiv keep the bareabs/<id>URL and record the paper id. - Flag uncited claims: any section making benchmark/experiment claims with NO link gets
flagged inline in the staged copy (
Citation: NONE PROVIDED…) and markedingest: BLOCKEDin the manifest — never fetch-and-ingest a paper you had to guess the identity of. - Dedup against the store per paper id (search content + provenance) — papers already
present are marked
skip: in store; hash-dedup only catches identical files, not the same paper cited by two digests. - Write
citations.jsonbeside the staged report: section, title, id, canonical URL, digest status (new vs continuing), ingest decision. This is the deterministic provenance for any follow-up per-paper ingest. - Ingest decisions: papers the digest marks new/highest-relevance (or the operator names) → fetch into the same staging dir and ingest individually. Continuing/"tracked" papers → do NOT bulk-ingest; note whether per-paper evidence exists in the store and surface the gap to the operator instead.
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.
- 4d ago First seen · 69 lines · 76 tokens per session scan A 50f8dd3e47fb
midmem-ingest is a skill published in the GitHub repository U4CSolutions/MidMem (2 stars, last pushed 10d ago), licensed Apache-2.0. It adds 76 tokens to every session and 1,152 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 skills, from other repositories
lemmalog
Externalize working memory and logical state into the lemmalog Datalog engine (MCP). Use for ANY multi-step task where state should outlive one context window or span agents: long investigations, debugging sessions, audits, multi-agent searches, systematic explorations, planning with many interdependent constraints…
autograph
Schema-as-code enforcement for any Obsidian vault. Zero hardcoded domains. Use when creating vault cards, checking vault health, running schema compliance, deduplicating entities, generating MOC indexes, running decay cycles, bootstrapping a vault, fixing wikilinks, finding orphans or backlinks, extracting entities…
dbrain-processor
Personal assistant for processing daily voice/text entries from Telegram. Classifies content, saves thoughts to Obsidian with wiki-links, generates HTML reports. Integrates Your Business context (clients, projects, CRM). Triggers on /process command or daily 21:00 cron.
ori-memory
Persistent agent memory with learning retrieval. Knowledge graph on markdown files — capture insights, decisions, research, and learnings during work, then retrieve them weeks or months later. Use when knowledge is too valuable to lose but too much to inject into every prompt.
braindb
Memory recall and save. Use at every conversation start and whenever the user shares personal information, expertise, project context, preferences, or decisions worth remembering long-term.
braindb-agent
Persistent memory across sessions via the BrainDB agent. Use at conversation start and whenever you need to recall what you know about the user or save new information to long-term memory.