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/juspay/neurolink/background-commandsgit clone --depth 1 https://github.com/juspay/neurolinkWhat 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 | $0.02727 |
| Opus 5 | $0.00000 | $0.01363 |
| Sonnet 5 | $0.00000 | $0.00545 |
| Haiku 4.5 | $0.00000 | $0.00273 |
Grade A, and why
BACKGROUND-COMMANDS scanned grade A with 1 finding 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 3d 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.
Runs shell commandslowCapability
Expected in a hook, worth knowing in a rule or an instructions file.
output at 100 KB; a `child_process` call with a command string is a shell How it starts
The opening of the file, as written. The whole thing — 225 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Background Commands (run_command_bg / command_status / command_output / command_kill)
A long-running agent has to run real commands — a build, a test suite, a linter
whose output is the evidence for a finding. The two obvious shapes both
fail: bashTool blocks the loop, hands the model a shell, and truncates its own
output at 100 KB; a child_process call with a command string is a shell
injection with extra steps.
These tools keep three promises instead.
| Promise | What it means |
|---|---|
| Detached | run_command_bg returns a taskId at once; the agent keeps working and asks about the command when it wants to. |
| Nothing discarded | Both streams go straight to files as they arrive, and the COMPLETE files are banked as artifacts when the command settles. The conversation gets a bounded tail plus a read-back call. |
| Hardened by contract | argv arrays with no shell, an exact-match executable allowlist, a realpath cwd sandbox, and a timeout that escalates SIGTERM → SIGKILL. |
Opt-in and additive: nothing registers these tools until you ask, and the policy is required — until one is set, every start is refused.
Quick start
import { NeuroLink } from "@juspay/neurolink";
const neurolink = new NeuroLink();
neurolink.registerBackgroundCommandTools({
allowedExecutables: ["/usr/local/bin/pnpm", "/usr/bin/git"],
cwdRoot: "/srv/checkout",
defaultTimeoutMs: 300_000,
maxOutputBytes: 32 * 1024 * 1024,
});
neurolink.registerTaskTools(); // optional — carries the command counters
await neurolink.generate({
input: {
text:
"Run the lint and test suites with run_command_bg, keep reviewing while " +
"they run, and read their output when command_status says they finished.",
},
maxSteps: 40,
});
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.
- 3d ago First seen · 225 lines · 0 tokens per session scan A 7e8b839e03a4
BACKGROUND-COMMANDS is an agent published in the GitHub repository juspay/neurolink (128 stars, last pushed 3d ago), licensed MIT. It costs nothing until one of its globs matches a file; then it loads 2,727 tokens. A static security scan graded it A with 1 finding (runs shell commands). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other agents, from other repositories
claude-design-orchestrator
Parses claude.ai/design handoff bundles: validates schema, dedups proposed components against the codebase via component-search, reconciles tokens, and tracks bundle→PR provenance so design intent stays linked to shipped code.
accessibility-specialist
Accessibility expert: WCAG 2.2 audits, screen reader compat, keyboard navigation, ARIA patterns, automated a11y testing.
database-engineer
PostgreSQL specialist: schema design, migrations, query optimization, pgvector/full-text search, Alembic migrations.
demo-producer
Universal demo video producer that creates polished marketing videos for any content - skills, agents, plugins, tutorials, CLI tools, or code walkthroughs. Uses VHS terminal recording and Remotion composition.
emulate-engineer
Stateful API emulation via Vercel emulate. Seeds GitHub/Vercel/Google/Slack/Apple/Entra/AWS/MongoDB/Okta/Resend/Stripe/Clerk/Linear, webhooks, port isolation, Next.js adapter. Use to replace flaky API mocks.
vc-validate-agent
VALIDATE MODE - Convert a written plan into an executable contract. Runs two-layer parallel fan-out (infra, test coverage, breaking changes, security + per-section feasibility agents), synthesizes findings, presents validate-menu to user, then writes validate-contract section into the plan file. Mandatory phase…