physical-capability-cloud CLAUDE.md

A set of Claude Code instructions for the physical-capability-cloud project. It requires checking and, when available, using a remote DGX Spark computer for heavy work and running security checks on new tools and dependencies.

In plain words
What is it for?
Use it when building, testing, installing dependencies, or vetting new packages and external tools in this project.
Why use it?
It sets project-specific rules for avoiding local memory problems and screening software for vulnerabilities, secrets, malware, and other risks.

Instructions file

Install

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.

agentmods
npx agentmods add instructions/lamasu/physical-capability-cloud/claude-md
Clone the repo
git clone --depth 1 https://github.com/LamaSu/physical-capability-cloud
Per session 12,990 This file is loaded in full into every session.
When invoked 12,990 The same file — it is already loaded in full.
Security scan C 4 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

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.

ModelPer sessionOnce invoked
Fable 5 $0.12990 $0.12990
Opus 5 $0.06495 $0.06495
Sonnet 5 $0.02598 $0.02598
Haiku 4.5 $0.01299 $0.01299

Measured 2d ago against content hash 5124d039a507, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade C, and why

physical-capability-cloud CLAUDE.md scanned grade C with 4 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.

Downloads and executes remote codemediumSupply chain

curl | sh runs whatever the server returns today, which is not necessarily what it returned when this was reviewed.

Dangerous keywords (rm -rf, git reset --hard, git push --force, drop table, curl | sh) are scored 0.0-1.0. Actions scoring >0.5 are dual-logged and may be denied per agent allowlist.

Downgraded: this mod is about security review, or the phrase is quoted, so it is likely naming the pattern rather than instructing it.

Reads agent configuration directorieslowAgent snooping

.claude/, .codex/, .gemini/ hold keys, settings and other credentials a mod has no legitimate need for.

**Configuration** (add to Claude Code settings or `~/.claude/settings.json`):

Downgraded: this mod is about security review, or the phrase is quoted, so it is likely naming the pattern rather than instructing it.

Recursive force deletemediumDestructive command

rm -rf with a variable or a broad path is one typo away from removing the wrong tree.

Dangerous keywords (rm -rf, git reset --hard, git push --force, drop table, curl | sh) are scored 0.0-1.0. Actions scoring >0.5 are dual-logged and may be denied per agent allowlist.

Downgraded: this mod is about security review, or the phrase is quoted, so it is likely naming the pattern rather than instructing it.

Makes network callslowCapability

Not a fault in itself. Listed so you know the mod talks to something, and to what.

Dangerous keywords (rm -rf, git reset --hard, git push --force, drop table, curl | sh) are scored 0.0-1.0. Actions scoring >0.5 are dual-logged and may be denied per agent allowlist.
CLAUDE.md · 1,048 lines

How it starts

The opening of the file, as written. The whole thing — 1,048 lines — stays where its author put it; the contents beside it link to each section on GitHub.

MANDATORY: DGX Spark Offload

ALWAYS check Spark availability (spark-check) at session start and offload ALL builds, tests, and heavy work to the DGX Spark (192.168.108.72, 119GB RAM). The local machine has only 16GB RAM and WILL OOM on builds, parallel tests, and multi-agent workloads. Use spark-run for everything. No exceptions.

  • spark-run "cd ~/projects/physical-capability-cloud && pnpm build --concurrency=1" — build
  • spark-run "cd ~/projects/physical-capability-cloud && pnpm --workspace-concurrency=1 -r test" — test
  • spark-run "cd ~/projects/physical-capability-cloud && pnpm install" — install
  • If Spark is down, use --concurrency=1 locally and warn the user about OOM risk.

MANDATORY: Security — Gate A Vetting Pipeline

All new tools, MCP servers, npm packages, and external dependencies MUST pass through Gate A vetting before being approved for use in this project.

How to Vet

  • Run /vet <package-path> on any new tool/package before integration
  • Runs up to 7 scanners: Trivy (vulns+SBOM), Gitleaks (secrets), ClamAV (malware), npm audit, pip-audit, Semgrep (SAST), prompt injection detection
  • Scanners degrade gracefully if not installed. Prompt injection scanner always runs.

Policy Thresholds (from ~/.claude/plugins/vetting-policy.json)

  • Critical vulnerabilities: 0 allowed (auto-reject)
  • High vulnerabilities: max 2 (warn on any)
  • Medium vulnerabilities: max 10
  • Secrets in source: 0 allowed (auto-reject)
  • Malware: auto-reject, no override
  • Prompt injection signals: max 1 (warn on any)

Verdicts

  • PASS: Clean, auto-approved
  • WARN: Suspicious findings, human reviews before use
  • FAIL: Critical vulns, secrets, or malware — auto-rejected

When to Vet

  • Before adding any new npm dependency to any package
  • Before integrating any MCP server
  • Before running any third-party script
  • After forging a new MCP server with /forge
  • Reports saved to ai/supervisor/forge_approvals/

MANDATORY: Action Classification

Read the full file on GitHub · 1,048 lines

Changes

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.

  1. 2d ago First seen · 1,048 lines · 12,990 tokens per session scan C 5124d039a507

Subscribe to this mod's changes

physical-capability-cloud CLAUDE.md is an instructions file published in the GitHub repository LamaSu/physical-capability-cloud (1 stars, last pushed 4d ago), licensed Apache-2.0. It adds 12,990 tokens to every session, about $0.0649 per session on Opus 5. A static security scan graded it C with 4 findings (downloads and executes remote code, reads agent configuration directories, recursive force delete). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.