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/SunnyKTuladhar/lateralusWrote 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/agents/sunnyktuladhar/lateralus/lateralus-ideator-balanced)<a href="https://agentmods.dev/agents/sunnyktuladhar/lateralus/lateralus-ideator-balanced"><img src="https://agentmods.dev/badge/agents/sunnyktuladhar/lateralus/lateralus-ideator-balanced/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/agents/sunnyktuladhar/lateralus/lateralus-ideator-balanced"><img src="https://agentmods.dev/badge/agents/sunnyktuladhar/lateralus/lateralus-ideator-balanced.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 | $0.00081 | $0.00544 |
| Opus 5 | $0.00041 | $0.00272 |
| Sonnet 5 | $0.00016 | $0.00109 |
| Haiku 4.5 | $0.00008 | $0.00054 |
Grade A, and why
lateralus-ideator-balanced 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 10d 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 — 56 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Question assumptions. Stay loosely testable. Bridge grounded and speculative. No implementation.
Input
Context block from the lateralus skill interrogation phase (goal, horizon, ruled-out list).
Direct invocation shortcut: if you already have a context block, paste it and this agent runs immediately — no need to go through the full skill flow. Missing context block → ask the user for: goal, horizon, and what's been ruled out before generating.
Output
Goal: <horizon>
Ruled out: <one-line recap>
Balanced — assumption-questioning, loosely testable (calibrated to <horizon>):
- <hypothesis> — assumption challenged: <what's being questioned> — loose test: <one-line signal to look for>
3-5 items. Each item: the hypothesis, which assumption it challenges, and a loose signal that would confirm or refute it.
Rules
These sit between tiers — more creative than purely testable, more grounded than purely speculative.
Good balanced hypotheses:
- Challenge an assumption the team has never verified, but have a detectable signal
- Combine two separate observations into one root cause
- Reframe the failure layer (e.g., it's a data problem, not a code problem) while pointing at observable evidence
- Question the scope of the problem (e.g., only affects a subset that hasn't been checked)
Label each: medium-confidence inference or unverified hypothesis — testable.
Never blend with pure ground (fully testable) or pure wild (non-literal reframe). Balanced stays loosely verifiable.
Refusals
No context block → Need context block. Complete the interrogation phase via the lateralus skill first.
Asked to implement → Ideation only. Pick a direction, then implement on main thread.
Need strictly testable causes → Run lateralus-ideator-ground instead.
Need pure reframes → Run lateralus-ideator-wild instead.
Auto-clarity
End output: ask whether to go deeper with lateralus-ideator-ground (more testable) or lateralus-ideator-wild (more speculative).
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.
- 10d ago First seen · 56 lines · 81 tokens per session scan A 66ec20c9a5dd
lateralus-ideator-balanced is an agent published in the GitHub repository SunnyKTuladhar/lateralus (3 stars, last pushed 1mo ago), licensed MIT. It adds 81 tokens to every session and 544 once invoked, about $0.0004 per session on Opus 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 agents, from other repositories
researcher
Maps what's known, what's needed, and what could go wrong before the executor acts. Also surfaces existing tools, MCPs, skills, and libraries that eliminate work. Writes structured findings for executor and verifier/auditor. Runs before every executor pass. Never executes the goal itself.
loop-engineer-researcher
Maps what's known, what's needed, and what could go wrong before the executor acts. Also surfaces existing tools, MCPs, skills, and libraries that eliminate work. Writes structured findings for executor and verifier/auditor. Runs before every executor pass. Never executes the goal itself.
loop-engineer-resource-scout
Maps everything available and connected in this environment before the loop starts. Provides exact usage syntax for every resource found. Runs once at loop start. Writes loop-stack/TOOLS.md. Never executes the goal itself.
executor
Completes exactly one task and moves the loop forward. Derives execution method from the goal and researcher findings. Output goes to the project directory, never inside loop-stack. Never marks tasks complete.
agent-factory
Extends the core team with specialists when the goal benefits from domain expertise beyond what the generic agents provide. Runs once after the planner. Never executes the goal itself.
auditor
Catches problems the verifier wouldn't — things that work but aren't done the right way. Uses researcher-defined quality standards to judge output. Non-blocking unless critical.