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.
git clone --depth 1 https://github.com/xzawed/claude-grok-build-pluginWrote 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/xzawed/claude-grok-build-plugin/memory)<a href="https://agentmods.dev/commands/xzawed/claude-grok-build-plugin/memory"><img src="https://agentmods.dev/badge/commands/xzawed/claude-grok-build-plugin/memory/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/commands/xzawed/claude-grok-build-plugin/memory"><img src="https://agentmods.dev/badge/commands/xzawed/claude-grok-build-plugin/memory.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 | $0.00006 | $0.00913 |
| Opus 5.5 | $0.00002 | $0.00365 |
| Sonnet 5.5 | $0.00001 | $0.00183 |
| Haiku 4.5 | $0.00001 | $0.00091 |
Grade A, and why
memory 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 14d 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
Call grok_cli with args starting ["memory", ...]. memory clear is the only
subcommand grok exposes — there is no list/search (grok memory list exits 2 with
unrecognized subcommand), so do not invent one. Memory is otherwise driven from grok's
own TUI (/memory …).
memory clear is destructive and asks Are you sure? [y/N] before deleting. This wrapper
gives grok no stdin, so an unanswered prompt returns Cancelled. immediately rather than
hanging — but that also means the clear does nothing. Confirm the scope with the user
here, then pass -y, which is what actually performs the deletion:
["memory","clear","-y"]— workspace memory (the default). grok picks the workspace from the git repository incwd(by itsorigin), so every clone and worktree of that repository shares it: pass the user's project folder ascwd.["memory","clear","--global","-y"]— global memory["memory","clear","--all","-y"]— both
When you confirm the scope — before the user's yes — say that each scope can go as a whole
folder: every note in it, its MEMORY.md, topics/, observations/, archive/, index
databases and, in the older layout, sessions/ — and, for the workspace, that it belongs to the
repository, so it goes for every clone and worktree of it. Say this whatever the grok version or
memory settings: understating a permanent deletion is the failure to avoid, and the narrower cases below
are not worth relying on. What was measured:
- grok 1.0.50 (2026-10-12), under every memory setting tried: the workspace clear deleted
<grok home>/memory-v2/workspaces/<project>-<hash8>/and--globaldeleted<grok home>/memory-v2/global/, each a whole folder;--allboth. The older<grok home>/memory/was left in place. - grok 1.0.44: the same folders with
[memory_v2] enabled = truein grok's config (the older tree stayed); without it, the older tree — the workspace's whole<grok home>/memory/<project>-<hash8>/folder (itsMEMORY.md,sessions/andindex.sqlite), and for--globalonly<grok home>/memory/MEMORY.md.
After a run with -y, relay what grok printed: its Cleared: lines are what went. grok prints
The following will be deleted: on a cancelled run too, so that list alone does not mean anything
went — Cancelled. and no Cleared: line mean nothing changed. A scope you asked for that grok
did not list (with --all, the workspace) was not cleared. Nothing to clear means only that
grok found nothing in the folders it targets, not that no memory is left: measured, grok 1.0.50
said it while notes sat in the older <grok home>/memory/, which it never clears. So do not tell
the user their memory is gone; for the workspace, check that cwd was the user's project, since
another repository's memory is a different workspace.
Never send memory clear — with or without -y — unless the user asked for a clear and you have
confirmed the scope. Without -y, today's grok asks and cancels, but that prompt is grok's, not a
safety net: do not send it without -y to see or list what would go, because a grok that stops
asking deletes on that run.
Present stdoutTail; on error show stderrTail. If status is blocked, relay the
message. Note the reported billing (the configured mode, not an observed charge).
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.
- 14d ago Changed · +19 lines 79f84e9ccc17
- 17d ago Changed · +8 lines 4f2e7607aeba
- 1mo ago Changed · +14 lines c93251df76bf
- 1mo ago First seen · 12 lines · 6 tokens per session scan A 5fba813a9635
memory is a command published in the GitHub repository xzawed/claude-grok-build-plugin (3 stars, last pushed today), licensed MIT. It adds 6 tokens to every session and 913 once invoked, about $0.0000 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 commands, from other repositories
recall-save
Generate / overwrite .recall/context.md with Recall's local offline summarizer.
simplify
The over-engineering review: five tags (delete, stdlib, native, yagni, shrink), a mandatory replacement per finding, and a real null result when there is nothing to cut.
git
The pre-finish status: branch, hygiene findings, message checks, workflow lint, template state.
test
Run the repository's actual test suite: every ecosystem's canonical runner — NOT run is never green.
ai-act-ask
Answer an EU AI Act question grounded in the bundled knowledge base — verbatim statute text, obligation paraphrases, and the compound-risk taxonomy. Offline and deterministic by default; cites the articles it relies on.
broadcast
Run broadcast on a distilled Raw — update related existing pages conversationally.