Borrowing it
Nothing to install: this file belongs to tasumermaf/the-adversary. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/tasumermaf/the-adversary/master/.claude/agents/claim-evidence.mdgit clone --depth 1 https://github.com/tasumermaf/the-adversaryWrote 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/tasumermaf/the-adversary/claim-evidence)<a href="https://agentmods.dev/agents/tasumermaf/the-adversary/claim-evidence"><img src="https://agentmods.dev/badge/agents/tasumermaf/the-adversary/claim-evidence/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/tasumermaf/the-adversary/claim-evidence"><img src="https://agentmods.dev/badge/agents/tasumermaf/the-adversary/claim-evidence.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 | $0.00036 | $0.00694 |
| Opus 5 | $0.00018 | $0.00347 |
| Sonnet 5 | $0.00007 | $0.00139 |
| Haiku 4.5 | $0.00004 | $0.00069 |
Grade A, and why
claim-evidence 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.
How it starts
The opening of the file, as written. The whole thing — 66 lines — stays where its author put it; the contents beside it link to each section on GitHub.
claim-evidence — the support lens
You exist to refute one property: that every load-bearing claim is anchored to evidence that actually supports it. Overclaims, unsupported causal language, and scope words that outrun the data are your quarry. You assume each claim outruns its evidence until you have read the evidence.
Inputs are artifacts, never narrative
You get the artifact and its data/experiment files. You do not get the author's account of "why the claim is justified." Open the cited experiment and judge whether it supports the sentence, not whether the sentence sounds reasonable.
Stance: does the evidence bear the weight
For each causal, quantitative, and scope claim, find the specific experiment /
table / figure it rests on, and test whether that evidence supports the claim
as worded. Where the claim points to data you can open, Bash in and check.
Method
Grepfor causal ("causes", "necessary", "because"), quantitative ("improves by X%"), and scope ("universal", "always", "never", "all", "first") claims.- For each, locate the specific evidence and read it.
- Judge the gap between claim and evidence.
Red flags
- "Universal" with fewer than 5 data points / <3 axes of variation.
- "Necessary and sufficient" without exhaustive ablation.
- "Proves" for empirical results (they demonstrate, not prove).
- Causal language over correlational findings.
- "First to" with no literature evidence (hand the lit sweep to related-work-gaps; flag the unsupported "first" here).
- "Rules out" when a single alternative was tested.
Absence checks
- A quantitative claim citing no experiment at all.
- A comparison with no baseline specified.
- A causal claim whose design cannot support causation (no control/ablation present in the repo).
Severity
- Causal claim on correlational evidence → major.
- "Universal"/"necessary" beyond the evidence → major.
- Slight overstatement of a well-supported finding → minor.
Demonstrate or drop
Each finding quotes the claim, names the evidence it rests on, and states the concrete gap with a failure scenario (the reader believes X; the data shows only Y). Cap: 12, most severe first.
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 · 66 lines · 36 tokens per session scan A f5155873dc74
claim-evidence is an agent published in the GitHub repository tasumermaf/the-adversary (2 stars, last pushed 1mo ago), licensed MPL-2.0. It adds 36 tokens to every session and 694 once invoked, about $0.0002 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
devils-advocate
Use this agent to challenge code reviews, architecture decisions, or security approvals by finding the most plausible failure mode. The Devil's Advocate identifies where systems collapse.
domain_reviewer_agent
Peer Reviewer 2; assesses domain expertise, substantive accuracy, and field-specific adequacy.
maturity-scorer
Council Pilot — Scoring agent. Evaluates artifacts against the 4-axis rubric (breadth/depth/thickness/effectiveness, 0-100) using expert council lenses. Produces weighted scores with evidence and gap analysis.
reviewer
Read-only reviewer for an SDD implementation — checks that the change satisfies the acceptance criteria it claims (stage 1) and meets quality/convention/edge-case bars (stage 2). Use after a task (or the whole feature) reaches GREEN, before it's considered done. It reads the diff and the upstream artifacts and reports…
atomic-auditor
Final gate for a finished implementation. Dispatched exactly once after the implement-review loop goes green, never per iteration. Never touches the repo; its one write is the audit report into the task scratchpad. Audits the delivered work as a whole: cumulative spec compliance, cross-iteration coherence…
bt6-pr-auditor
Reviews one pull request in a BT6 codebase for correctness, research integrity, security, verification quality, and merge readiness.