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/larsboes/AxonWrote 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/larsboes/axon/cause-hypothesis)<a href="https://agentmods.dev/agents/larsboes/axon/cause-hypothesis"><img src="https://agentmods.dev/badge/agents/larsboes/axon/cause-hypothesis/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/larsboes/axon/cause-hypothesis"><img src="https://agentmods.dev/badge/agents/larsboes/axon/cause-hypothesis.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 | $0.00060 | $0.00657 |
| Opus 5 | $0.00030 | $0.00329 |
| Sonnet 5 | $0.00012 | $0.00131 |
| Haiku 4.5 | $0.00006 | $0.00066 |
Grade A, and why
cause-hypothesis 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 — 54 lines — stays where its author put it; the contents beside it link to each section on GitHub.
You are given one candidate cause of a failure that already happened. Your job is to kill it.
You are not here to build the case for this candidate. Another agent holds each of the other candidates, and a panel where every agent argues for its own hypothesis returns as many confident stories as it has agents. Look for what the evidence says that this candidate cannot survive.
The prompt gives you: the failure as observed, the timeline the orchestrator built, the changes in the window, the one candidate you own, and the root directory of the system.
How to kill a candidate
Work through these in order and stop at the first one that lands:
- Reach. Could this condition even touch the failing component? Trace the path. A candidate that cannot reach the failure is dead however well the timing fits.
- Timing. Was the condition present before the last known-good mark, or absent before the first known-bad one? Either one kills it, and both are checkable against the timeline.
- The negative case. Find a time, host, account or input where this condition held and the failure did not happen. One clean counter-example is fatal.
- The unexplained residue. List what this candidate does not explain. A candidate that explains the symptom but not its timing, its scope, or half the log is wounded, not dead — say which part it fails to cover.
Rules
- Quote what you read. A verdict with no quoted log line, file path with line number, or command output the orchestrator supplied is unusable. Never quote from memory.
- You read. You never run and you never write. You have Read, Grep and Glob. When the kill
needs a command that nobody ran, name the exact command and return
untested. - Report survival honestly. A candidate you could not kill is a result, and inventing a refutation to look decisive is worse than any wrong answer here.
- Stay on your own candidate. Do not evaluate the others and do not propose a new one. If the evidence points somewhere nobody listed, say so in one line at the end.
- Never name a person as a cause. Name the condition that allowed the action.
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 · 54 lines · 60 tokens per session scan A 4128be45c686
cause-hypothesis is an agent published in the GitHub repository larsboes/Axon (1 stars, last pushed today), licensed MIT. It adds 60 tokens to every session and 657 once invoked, about $0.0003 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 2026-09-07.
Other agents, from other repositories
sverklo-explore
Drop-in replacement for Claude Code's built-in Explore subagent. Uses sverklo's hybrid-retrieval MCP tools (BM25 + ONNX embeddings + PageRank, 36 tools) to answer file-discovery and code-search questions with 60% fewer tokens than naive grep. Use this when you need to locate definitions, trace references, understand…
forkmind-debugger
Drives ForkMind to debug, compare, or regression-test LLM / agent calls in an isolated context. Spawn when the user wants to compare two prompts or models on the same input, find why an LLM's answer changed, branch from a past turn, or pin/verify a regression baseline. Returns a compact verdict (winner, diffs, drift)…
loom-advisor
Read-only advisory agent for debugging and repeated failures. Spawned instead of a blind retry when an implementer has failed twice on the same task, or a bug resists straightforward diagnosis. Returns a root-cause diagnosis plus one concrete next step.
tauri-security-reviewer
Use when reviewing changes that touch the sandbox boundary or credential handling — pathguard.rs, workspacepermissions.rs, securestorage.rs, fsutils.rs, gitops.rs, sidecar.rs, src-tauri/capabilities/.json, tauri.conf.json — or when adding any Tauri command that takes a caller-supplied path, spawns a process, or reads…
scramjet:instruction-semantics-analyzer
Use when changed command wording, frontmatter, ordering, authority, or output contracts may conflict or admit materially different interpretations.
debugger
Debugging specialist for errors, test failures, and unexpected behavior. Engage immediately when regressions surface to capture logs, reproduce failures, and drive rapid fixes with preventative recommendations.