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 agents/yogasw/wick/memory-guardgit clone --depth 1 https://github.com/yogasw/wickWhat 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 | $0.00000 | $0.03411 |
| Opus 5 | $0.00000 | $0.01706 |
| Sonnet 5 | $0.00000 | $0.00682 |
| Haiku 4.5 | $0.00000 | $0.00341 |
Grade A, and why
memory-guard 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.
How it starts
The opening of the file, as written. The whole thing — 170 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Memory Guard
The agent pool (Pool & Sessions) caps how many agent processes run at once. It has no idea how much RAM any one of them uses — an idle agent at ~150 MB and one driving a browser at ~2 GB count as the same slot. A single runaway agent (a browser leak, a bad loop) can take the whole machine down with it, wick included.
The memory guard adds a second axis: byte limits, enforced by the kernel, per agent. It ships off — a fresh install and an upgraded one behave identically until you opt in.
::: info Source
Config + arithmetic: internal/agents/config/memguard.go.
Mechanism: internal/agents/provider/memscope/ (systemd scopes and the raw-cgroupfs fallback), internal/agents/provider/memguard.go (wiring into spawn).
Measurement: internal/pkg/memreport/ (live /proc reads + history buffer).
:::
Linux only
Enforcement needs a Linux kernel. Off Linux — Windows, macOS — the guard degrades cleanly: no limit is applied, and both the CLI and the Resources page say so plainly instead of silently doing nothing.
On Linux, wick probes two mechanisms in order at first use and caches whichever works, by actually creating and tearing down a throwaway group rather than guessing from paths or permissions:
| Backend | Requires | Notes |
|---|---|---|
| systemd (preferred) | a reachable systemd user session (systemd-run --user --scope) |
Scopes are reaped automatically (--collect), and cgroup v2's memory.events gives a real per-scope OOM-kill counter. |
| cgroupfs (fallback) | a writable /sys/fs/cgroup/memory (cgroup v1) |
No systemd needed — the kernel is enough. Used on hosts whose PID 1 isn't systemd (a Fly.io Machine, a bare container, Termux without linger). |
| none | — | Agents run unguarded; only /proc measurement remains. |
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 · 170 lines · 0 tokens per session scan A cc23c068e246
memory-guard is an agent published in the GitHub repository yogasw/wick (5 stars, last pushed 4d ago), licensed MIT. It costs nothing until one of its globs matches a file; then it loads 3,411 tokens. 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 agents, from other repositories
desktop
Agent "desktop" from WrongStack/WrongStack, covering working rules and output.
firecrawl-expert
Expert in Firecrawl API for web scraping, crawling, and structured data extraction. Handles dynamic content, anti-bot systems, and AI-powered data extraction.
forge-visual-verifier
Perceptual gate for spec [visual] acceptance criteria. Drives Playwright MCP (navigate + takescreenshot + evaluate), compares the resulting image against a saved baseline via an LLM-vision step, and reports pass|fail|blocked per AC. Invoked after all task-level structural checks pass and before FORGECOMPLETE is…
project-manager
Project manager for CrawlForge MCP Server development. Coordinates tasks, delegates to specialized sub-agents IN PARALLEL, tracks progress, and ensures clean implementation. Use PROACTIVELY for any multi-step project coordination.
mcp-implementation
MCP server implementation specialist. Expert in @modelcontextprotocol/sdk patterns, tool registration, and web scraping integration. Use for implementing server code, tools, and core functionality.
security-auditor
Security specialist for CrawlForge MCP Server. Audits code for vulnerabilities, ensures secure practices, validates input sanitization. Use PROACTIVELY before deployments and after major changes.