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 agentmods add skills/getsentry/warden-skills/vercel-deepsecnpx skills add getsentry/warden-skills --skill vercel-deepsecgit clone --depth 1 https://github.com/getsentry/warden-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/getsentry/warden-skills/vercel-deepsec)<a href="https://agentmods.dev/skills/getsentry/warden-skills/vercel-deepsec"><img src="https://agentmods.dev/badge/skills/getsentry/warden-skills/vercel-deepsec.svg" alt="Measured on agentmods" 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 | $0.00090 | $0.04392 |
| Opus 5 | $0.00045 | $0.02196 |
| Sonnet 5 | $0.00018 | $0.00878 |
| Haiku 4.5 | $0.00009 | $0.00439 |
Grade A, and why
vercel-deepsec scanned grade A with 2 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 6d 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.
Makes network callslowCapability
Not a fault in itself. Listed so you know the mod talks to something, and to what.
return requests.get(request.GET["url"], allow_redirects=True).text Runs shell commandslowCapability
Expected in a hook, worth knowing in a rule or an instructions file.
- Shell command strings built with user data: Python `shell=True`, `os.system`, Node `exec`, Go shell wrappers, Ruby backticks, PHP `system`, Lua `os.execute`. How it starts
The opening of the file, as written. The whole thing — 326 lines — stays where its author put it; the contents beside it link to each section on GitHub.
You are a senior application security researcher. This skill adapts the Vercel Labs DeepSec default processor prompt into Warden form. It is intentionally broad because it is used for benchmark and comparison runs.
For focused production review, prefer a narrower Warden skill when one maps directly to the concern. When this benchmark skill runs, keep the bar high: report only exploitable vulnerabilities with a traced source, sink, missing guard, impact, and fix.
Source provenance and benchmark notes live in SOURCES.md.
Benchmark Contract
- Use the same criteria across runs. Do not tune the analysis to expected answers.
- Treat scanner hits, grep hits, and suspicious filenames as starting points only.
- Investigate beyond the flagged pattern when the surrounding file exposes a different bug.
- Return no findings for generated, vendored, gitignored, build output, fixture-only, or non-production code unless the benchmark target explicitly includes it.
- Do not inflate severity to make a benchmark look better. A noisy high is worse than an empty result.
Trace. Do Not Skim.
The sink tells you what could happen. The source tells you whether it will.
- Read each target file fully before reporting.
- Identify attacker-controlled input: request body, query string, path params, cookies, headers, webhook payloads, OAuth callback params, uploaded files, user profile fields, database values written by users, third-party callbacks, or caller-controlled workflow/service inputs.
- Identify the security boundary: login state, tenant, team, org, project, account, role, OAuth state, webhook signature, internal network, filesystem root, cache namespace, or paid API quota.
- Follow imports, wrappers, middleware, guards, serializers, validators, route definitions, shared utilities, and sibling handlers.
- Verify mitigations in context. Parameterized queries, exact allowlists, safe URL fetchers, escaping, signature checks, handler-level auth wrappers, and resource-level ownership checks can close the path.
- Use
rgandgit log -p <file>when needed to compare sibling handlers or see whether a guard was recently removed. - Drop speculation. If the dataflow, boundary, or missing mitigation cannot be proven from available code, do not report it.
What ships with it
4 files 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.
- 6d ago First seen · 326 lines · 90 tokens per session scan A 3e1617debed1
vercel-deepsec is a skill published in the GitHub repository getsentry/warden-skills (57 stars, last pushed 14d ago), licensed MIT. It adds 90 tokens to every session and 4,392 once invoked, about $0.0005 per session on Opus 5. A static security scan graded it A with 2 findings (makes network calls, runs shell commands). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
django-perf-review
Django performance code review. Use when asked to "review Django performance", "find N+1 queries", "optimize Django", "check queryset performance", "database performance", "Django ORM issues", or audit Django code for performance problems.
django-access-review
Django access control and IDOR security review. Use when reviewing Django views, DRF viewsets, ORM queries, or any Python/Django code handling user authorization. Trigger keywords: "IDOR", "access control", "authorization", "Django permissions", "object permissions", "tenant isolation", "broken access".
gha-security-review
GitHub Actions security review for workflow exploitation vulnerabilities. Use when asked to "review GitHub Actions", "audit workflows", "check CI security", "GHA security", "workflow security review", or review .github/workflows/ for pwn requests, expression injection, credential theft, and supply chain attacks.…
presentation-creator
Create data-driven presentation slides using React, Vite, and Recharts with Sentry branding. Use when asked to "create a presentation", "build slides", "make a deck", "create a data presentation", "build a Sentry presentation". Scaffolds a complete slide-based app with charts, animations, and single-file HTML output.
triage-frontend-issues
Triage new issues in the Sentry javascript project by archiving non-actionable noise. Use when asked to "triage issues", "triage the javascript project", "archive non-actionable issues", "triage new frontend issues", or "clean up the sentry/javascript queue". Operates only on the sentry/javascript project, only…
sred-project-organizer
Take a list of projects and their related documentation, and organize them into the SRED format for submission.