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 AtlasOmnia/hermes-custom-pack --skill evidence-based-repliesgit clone --depth 1 https://github.com/AtlasOmnia/hermes-custom-packWrote 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/atlasomnia/hermes-custom-pack/evidence-based-replies)<a href="https://agentmods.dev/skills/atlasomnia/hermes-custom-pack/evidence-based-replies"><img src="https://agentmods.dev/badge/skills/atlasomnia/hermes-custom-pack/evidence-based-replies/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/atlasomnia/hermes-custom-pack/evidence-based-replies"><img src="https://agentmods.dev/badge/skills/atlasomnia/hermes-custom-pack/evidence-based-replies.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 | $0.00043 | $0.00816 |
| Opus 5 | $0.00022 | $0.00408 |
| Sonnet 5 | $0.00009 | $0.00163 |
| Haiku 4.5 | $0.00004 | $0.00082 |
Grade A, and why
evidence-based-replies 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 9d 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.
This is a copy
100% identical to evidence-based-replies — 0 lines differ, which has more behind it and is treated as the original. This page carries a canonical link to it rather than competing with it.
How it starts
The opening of the file, as written. The whole thing — 86 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Evidence-Based Replies
Use this skill when the user wants to:
- compare someone's claim to a research paper, benchmark, article, or repo README
- draft a reply that corrects a misunderstanding of the evidence
- separate "the source supports X" from "the source does not support Y"
- turn a source-comparison into a Reddit/X/forum-ready response
Core principle
Do not stop at summarizing the source. The deliverable is usually the distinction:
- what the source actually establishes
- what stronger conclusion the other person is trying to import into it
- why that leap does not follow
If the user is asking for a reply, the misunderstanding of the evidence must be addressed explicitly, not left implicit.
Default workflow
- Read the conversation/claim carefully.
- Extract the specific proposition the other person is asserting from the source.
- Read the cited source.
- Split findings into:
- Supported by the source
- Not established by the source
- Reasonable but unproven extrapolations
- Compare the source type:
- empirical paper
- benchmark paper
- opinion/design README
- repo guidance / author doctrine
- Draft the response around the distinction, not around vague disagreement.
Required output shape
When explaining the comparison, prefer this structure:
- "The paper/source does show ..."
- "What it does not show is ..."
- "That is the step I disagree with."
- "So my criticism is not X; it is Y."
This pattern prevents straw-manning and keeps the correction precise.
Writing guidance for this user
the user prefers concise, practical responses. For forum replies:
- lead with the misunderstanding, not with throat-clearing
- avoid generic "both sides have a point" filler
- make the inferential gap explicit
- distinguish directionally true claims from overclaims
- prefer operational language over academic hedging
If he says a draft "needs to address the misunderstanding of the research," revise around the evidentiary gap immediately.
What ships with it
1 file beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
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.
- 9d ago First seen · 86 lines · 43 tokens per session scan A 81eebd0e3ff8
evidence-based-replies is a skill published in the GitHub repository AtlasOmnia/hermes-custom-pack (56 stars, last pushed 25d ago), licensed MIT. It adds 43 tokens to every session and 816 once invoked, about $0.0002 per session on Opus 5. A static security scan graded it A with 0 findings. It is 100% identical to evidence-based-replies, differing in 0 lines, and is treated as a copy.
Other skills, from other repositories
config
Use when configuring obshare-cli credentials, checking saved settings, testing connectivity, clearing local config, or setting optional Obsidian bridge values used by the V0.2.0 direct-share stack.
obshare-cli
Use when setting up obshare-cli 0.2.0, reviewing available commands, or choosing the right command for config, upload, permission, history, delete, and the plugin-first direct-share workflow.
agent-dreaming-agnostic
Memory-agnostic background consolidation — reviews sessions, scores candidates, promotes durable insights to the active memory backend (built-in MEMORY.md or Holographic). Three-phase (Light/Deep/Condensation/REM). Run via cron or manually.
upload
Use when uploading a Markdown note to Feishu, optionally applying public or copy/download permissions, or requesting JSON upload results for automation or the Obsidian plugin direct-share flow.
permission
Use when changing Feishu document visibility, copy permissions, or download permissions for an uploaded document token.
delete
Use when removing a Feishu document by token and cleaning the matching entry from local obshare-cli upload history.