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 skills add sruthir28/enterprise-ai-skills --skill hypothesis-treegit clone --depth 1 https://github.com/sruthir28/enterprise-ai-skillsWrote 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/skills/sruthir28/enterprise-ai-skills/hypothesis-tree)<a href="https://agentmods.dev/skills/sruthir28/enterprise-ai-skills/hypothesis-tree"><img src="https://agentmods.dev/badge/skills/sruthir28/enterprise-ai-skills/hypothesis-tree/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/skills/sruthir28/enterprise-ai-skills/hypothesis-tree"><img src="https://agentmods.dev/badge/skills/sruthir28/enterprise-ai-skills/hypothesis-tree.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector pass
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 | $0.00097 | $0.01979 |
| Opus 5 | $0.00048 | $0.00989 |
| Sonnet 5 | $0.00019 | $0.00396 |
| Haiku 4.5 | $0.00010 | $0.00198 |
Grade A, and why
hypothesis-tree 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 11d 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 — 151 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Hypothesis Tree
A hypothesis tree is your Day-1 best guess, structured so you can disprove it fast. Issue trees decompose the question; hypothesis trees commit to an answer and tell you what to go test.
The McKinsey discipline: don't start work without a hypothesis. Otherwise you research forever, find nothing surprising, and produce a "comprehensive overview" no one acts on.
Issue tree vs. hypothesis tree
| Issue Tree | Hypothesis Tree | |
|---|---|---|
| Top | The question | The answer (your guess) |
| Branches | Sub-questions (MECE problem space) | Sub-claims that, if true, prove the top claim |
| Bottom | Areas to investigate | Specific tests that would kill the claim |
| Use when | You don't know what matters yet | You have a strong prior and want to test it efficiently |
| Risk if skipped | You miss a dimension | You waste 6 weeks on analysis that doesn't move conviction |
Both belong in the consultant toolkit. Issue tree first (frame the space), hypothesis tree second (commit and test). Skipping the hypothesis tree is the #1 reason strategy work takes 3x longer than it should.
Design choices
- You must commit to a Day-1 answer. "I don't know" is not a hypothesis. Make a guess. Be wrong loudly and fast.
- 3 sub-hypotheses, not more. If you need 5, your top hypothesis isn't well-formed.
- Every leaf is a test that could kill the branch. "Do more research" is not a test. "Pull pricing data for 4 competitors and compare to ours" is a test.
- State conviction explicitly. Each sub-hypothesis gets a confidence (High / Med / Low). Tests focus on the low-conviction branches first — those are where you'll learn the most.
- Disconfirmation > confirmation. Design tests to kill the hypothesis, not validate it. If you can't think of what would disprove it, you're not testing — you're rationalizing.
Structure
TOP HYPOTHESIS (your Day-1 answer to the governing question)
│
├─ SUB-HYPOTHESIS 1 [confidence: H/M/L]
│ └─ Test: [Specific analysis, named source, kill-criterion]
│
├─ SUB-HYPOTHESIS 2 [confidence: H/M/L]
│ └─ Test: [Specific analysis, named source, kill-criterion]
│
└─ SUB-HYPOTHESIS 3 [confidence: H/M/L]
└─ Test: [Specific analysis, named source, kill-criterion]
PRIORITY: Test [low-conviction sub] first. Kill-criterion: [what result makes us abandon the top hypothesis].
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.
- 11d ago First seen · 151 lines · 97 tokens per session scan A 986ad10c7e05
hypothesis-tree is a skill published in the GitHub repository sruthir28/enterprise-ai-skills (139 stars, last pushed 9d ago), licensed MIT. It adds 97 tokens to every session and 1,979 once invoked, about $0.0005 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 2026-08-30.
Other skills, from other repositories
lyrik
Security assessment of a codebase — minimal mode for runner validation.
librarian
Read-only retrieval over the daily-digest kept set. Slash-invoke with /librarian ; returns items verbatim with citations.
notes
Create and manage text notes in the workspace.
tdx-assets
Look up staff devices, assets, and locations in TeamDynamix (TDX). Given a list of names, emails, or partial identifiers, resolve each person and return the assets assigned to them with location, model, ownership type, and status. Read-only.
aggregator
Daily fetch from a fixed public allowlist; score against the user's interests file; cluster into themes; push the digest to the configured channel.
file-search
Find files and search content across directories.