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 Kshitijpalsinghtomar/depth-skills --skill ds-adversarygit clone --depth 1 https://github.com/Kshitijpalsinghtomar/depth-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/kshitijpalsinghtomar/depth-skills/ds-adversary)<a href="https://agentmods.dev/skills/kshitijpalsinghtomar/depth-skills/ds-adversary"><img src="https://agentmods.dev/badge/skills/kshitijpalsinghtomar/depth-skills/ds-adversary/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/kshitijpalsinghtomar/depth-skills/ds-adversary"><img src="https://agentmods.dev/badge/skills/kshitijpalsinghtomar/depth-skills/ds-adversary.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 | $0.00022 | $0.01433 |
| Opus 5 | $0.00011 | $0.00717 |
| Sonnet 5 | $0.00004 | $0.00287 |
| Haiku 4.5 | $0.00002 | $0.00143 |
Grade A, and why
adversary 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 — 132 lines — stays where its author put it; the contents beside it link to each section on GitHub.
ADVERSARY — Self-Opposition Engine
You have an answer. Before you deliver it — write the case that it should be rejected.
Not a balanced review. Not "on the other hand." A prosecution. You are the most competent opponent this answer will ever face. Build the case for rejection with the same quality you used to build the answer.
The Failure Mode You Must Recognize
You are about to generate a review of your own work that:
- Lists objections you already know how to dismiss (shadowboxing)
- Uses softening language: "one might argue," "a potential concern is" (distancing from the attack)
- Grades every objection as "Minor" because nothing feels truly threatening to the answer you're already committed to
- Concludes "the approach is sound" without having genuinely tested whether it is
This is confirmation cascade — each supporting token makes the next supporting token more likely. The review becomes a rubber stamp. Breaking the cascade requires generating content that actively undermines your own conclusion.
The Protocol
1 — STATE: Write the Answer Under Review
One paragraph. What is the answer, recommendation, or plan you are about to deliver?
Write it clearly enough that an opponent could attack it. If you can't state it in one paragraph — the answer isn't coherent enough to review.
Artifact: The stated answer. Everything below attacks this specific text.
2 — ATTACK: Write Five Specific Attacks Against THIS Answer
Not generic concerns. Attacks on THIS specific answer for THIS specific problem.
For each attack, use this template:
ATTACK [N]: [one-line summary]
Claim: [the specific thing that is wrong, incomplete, or dangerous]
Evidence: [why this attack is plausible — cite specific aspects
of the answer, the domain, or the context]
If true: [what happens — the specific consequence]
Attack axis checklist — write at least one attack per axis:
- Correctness attack: "Step/claim X is factually wrong because [specific reason]." Not "might be wrong" — write it as if you believe it IS wrong.
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 · 132 lines · 22 tokens per session scan A ef88c30c9252
adversary is a skill published in the GitHub repository Kshitijpalsinghtomar/depth-skills (4 stars, last pushed 3mo ago), licensed MIT. It adds 22 tokens to every session and 1,433 once invoked, about $0.0001 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 skills, from other repositories
vox-video-director
Turn ONE topic into a finished Vox-style paper-collage explainer / ad video, end to end with Aliyun Bailian CLI + local ffmpeg — script, collage keyframes, motion, voice-over, music, captions, all automated. Use this whenever the user wants a "Vox style" video, a paper/torn-paper collage animation, a "motion collage"…
bailian-train-deploy
A workflow for using Alibaba Cloud’s Bailian command-line tool to fine-tune or directly deploy AI models as callable services. It covers text, speech-synthesis, image-generation, and video-generation models.
api-database-mongodb
Native MongoDB driver (the mongodb npm package) - MongoClient lifecycle, typed collections, CRUD result shapes, cursors, aggregation pipelines, index design, transactions.
ai-infrastructure-ollama
Local LLM inference with the Ollama JavaScript client -- chat, streaming, tool calling, vision, embeddings, structured output, model management, and OpenAI-compatible endpoint.
ai-orchestration-langchain
LangChain.js patterns for building LLM applications — chat models, LCEL chains, prompt templates, structured output, agents, tools, RAG, streaming, and LangSmith tracing.
ai-provider-cohere-sdk
Official Cohere TypeScript SDK patterns -- CohereClientV2, chat, embeddings, rerank, RAG with citations, tool use, streaming, and model selection.