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 skills add 3M1RY33T/loci --skill skillgit clone --depth 1 https://github.com/3M1RY33T/lociWrote 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/3m1ry33t/loci/skill)<a href="https://agentmods.dev/skills/3m1ry33t/loci/skill"><img src="https://agentmods.dev/badge/skills/3m1ry33t/loci/skill/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/3m1ry33t/loci/skill"><img src="https://agentmods.dev/badge/skills/3m1ry33t/loci/skill.svg" alt="Reviewed on agentmods" width="80" 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 5.1 | $0.00103 | $0.02326 |
| Opus 5.5 | $0.00041 | $0.00930 |
| Sonnet 5.5 | $0.00021 | $0.00465 |
| Haiku 4.5 | $0.00010 | $0.00233 |
Grade A, and why
loci 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 1mo 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 — 188 lines — stays where its author put it; the contents beside it link to each section on GitHub.
/loci
Scoped memory for coding agents. A router in front of two stores: a structure
store (what calls what, with file:line citations) and an episode store
(what happened and why — READMEs, docs, commit bodies, comment blocks). One
question path serves "how does auth work here" and "have I solved this in any
project"; only the size of the scope set changes.
Two properties shape everything below. Scopes are never merged — a question is routed to at most three projects and each is searched alone. Both layers can refuse — the router can decline to pick a project, and a project it picked can hand back nothing. A refusal is an answer.
Usage
/loci "<question>" ask memory; routes on your working directory
/loci update refresh everything registered
/loci update <root> ...and scan there for projects added since
/loci doctor coverage gaps per project, and the fix for each
/loci scopes what is registered, and the group each is in
/loci route "<question>" where a question routes, and why it abstained
/loci add <path> register one project (`/loci add .` for this repo)
/loci setup [dirs…] first run: scan, graph, index, embed, calibrate
/loci <anything else> passed through to the `loci` CLI verbatim
What to do when invoked
If the user typed /loci --help or /loci -h with nothing else, print the
Usage block above verbatim and stop.
If command -v loci finds nothing, say so once and stop: the fix is
pipx install 'loci-mem[all]', and no subcommand can work without it.
Then dispatch on the first word. A first word that is not a subcommand below and
not a flag is a question — send it to ask, do not treat it as a path.
Anything that names a real loci subcommand not documented here (index,
embed, graphs, calibrate, eval, groups, group, scan, mcp) is
passed through unchanged.
Prefer the MCP server when it is connected. If tools named ask, scopes
and doctor are available from a loci MCP server, call those instead of
shelling out, and pass cwd explicitly — the server holds the embedding model
and the largest projects' rankers in memory, so a call costs ~0.1s against the
~4.4s a cold CLI invocation pays in imports and model construction. Everything
else (update, route, add, setup) has no tool; use the CLI.
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.
- 1mo ago Changed · +16 lines 060ab84008bc
- 1mo ago Changed · +2 lines c228cee0f4d2
- 1mo ago First seen · 170 lines · 103 tokens per session scan A d1832f8ca6c1
loci is a skill published in the GitHub repository 3M1RY33T/loci (1 stars, last pushed 1mo ago), licensed MIT. It adds 103 tokens to every session and 2,326 once invoked, about $0.0004 per session on Opus 5.5. 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
post-cortex-memory
Long-term project memory through the post-cortex MCP server (tools named session, semanticsearch, updateconversationcontext, assemblecontext, queryconversationcontext, getstructuredsummary, manageworkspace, manageentity, admin). Use it whenever those tools are available and the work could benefit from what was learned…
rekal-init
Bootstrap rekal memory for a project. Scans the codebase for architecture, conventions, dependencies, workflows, and config, then stores durable knowledge as properly typed, tagged, deduplicated memories. Use when starting rekal on a new project, or when user says "init rekal", "bootstrap memory", "populate rekal"…
rekal-save
End-of-session memory capture with deduplication. Extracts durable knowledge, checks for duplicates, stores or replaces as appropriate. Use whenever a session wraps up, a task finishes, or the user says goodbye/thanks/done. Also use when significant preferences, decisions, or discoveries emerge mid-session. Make sure…
rekal-usage
Operational guide for rekal memory tools. Precise rules for when/how to call each tool, with exact parameters and decision trees. Use at session start, when onboarding to a rekal workspace, or when user asks "how do I use rekal", "what rekal tools", "help with memory". Trigger: /rekal-usage.
alice-memory
Use Alice as the user's durable local memory. Load before answering from context, and whenever you learn something worth keeping across sessions.
alice-project-memory
Use Alice as the project-scoped memory and continuity layer. Load before build or review work, and whenever a decision or constraint is worth keeping.