Decepticon is an autonomous red-team agent that coordinates AI agents, security tools, sandboxes, and supporting services for authorized cybersecurity assessments. Security researchers and red teams can run it through its Docker stack, cloud service, command-line interface, or Python SDK, with the catalogue entries representing its available skills.
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 PurpleAILAB/Decepticon --skill seven-question-gategit clone --depth 1 https://github.com/PurpleAILAB/DecepticonWrote 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/purpleailab/decepticon/seven-question-gate)<a href="https://agentmods.dev/skills/purpleailab/decepticon/seven-question-gate"><img src="https://agentmods.dev/badge/skills/purpleailab/decepticon/seven-question-gate/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/purpleailab/decepticon/seven-question-gate"><img src="https://agentmods.dev/badge/skills/purpleailab/decepticon/seven-question-gate.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector warn
SkillSpector: 1 finding, up to medium
These are SkillSpector’s own severities. On a checked sample its high-severity flags on skills were ~96% false positives — a documented command, a public API, a “never do X” rule — so we show them as a caution to read, not a verdict. Why →
- medium Agent Snooping · line 175 Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.Fix: Remove all code or instructions that list or read other skills' files or directories. Skills should operate independently; cross-skill access is a privilege escalation.
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 | $0.00039 | $0.01637 |
| Opus 5 | $0.00019 | $0.00818 |
| Sonnet 5 | $0.00008 | $0.00327 |
| Haiku 4.5 | $0.00004 | $0.00164 |
Grade A, and why
seven-question-gate 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 — 178 lines — stays where its author put it; the contents beside it link to each section on GitHub.
7-Question Gate
Run this gate after validate_finding returns success but BEFORE
adding the finding to the report. Any "no" → kill the finding,
don't write a report. This saves bounty validity-ratio and engagement-
report quality.
The 7 questions
1. Is the asset in scope?
Check the engagement's scope.md (recon/decepticon's RoE doc):
- In-scope domain list
- In-scope IP/CIDR list
- Out-of-scope explicit exclusions (test envs, staging, third-party CDNs/CDN-managed subdomains where the org doesn't own the underlying machine)
If asset is not on the in-scope list OR on the out-of-scope list: kill. Do not write the report.
2. Is there real-world impact?
Theoretical bugs without demonstrable impact get N/A on every BB program. Concrete impact statements include:
- "An attacker can read victim user X's PII (email, name, DOB, SSN)"
- "An attacker can post on behalf of victim user X"
- "An attacker can transfer funds from victim user X's account"
- "An attacker can persist code on the production server"
- "An attacker can pivot to internal network 10.0.0.0/8"
If the finding's impact is "configuration is non-default" or "the manual recommends X but the deployment does Y" without concrete attacker-reachable harm: kill.
3. Does the PoC actually prove the impact?
The validate_finding result is necessary but not sufficient. The
PoC must demonstrate the IMPACT, not just trigger the vector.
Examples:
- IDOR PoC must show ATTACKER session reading VICTIM data — not just "request returned 200 OK"
- XSS PoC must execute attacker-controlled JS in victim's browser context — not just "alert(1) reflected in HTML source"
- SQLi PoC must extract real data — not just "single quote → 500"
- SSRF PoC must reach an internal-only target — not just "external fetch worked"
If the PoC stops short of impact demonstration: kill or queue for re-verification with a better PoC.
4. Is the impact above the program's severity floor?
Many BB programs explicitly out-of-scope:
- CSRF on logout endpoint
- Self-XSS (requires victim to inject own payload)
- Missing security headers
- Information disclosure of public-by-design info
- Rate-limit issues w/o demonstrated abuse
- Subdomain takeover candidates where ownership can't be proven
- Clickjacking without authenticated state change
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 · 178 lines · 39 tokens per session scan A f97244c0537c
seven-question-gate is a skill published in the GitHub repository PurpleAILAB/Decepticon (5,463 stars, last pushed 9d ago), licensed Apache-2.0. It adds 39 tokens to every session and 1,637 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 2026-08-30.
Other skills, from other repositories
chinese-code-review
A Chinese-language reference for discussing code review findings with a development team.
receiving-code-review
A workflow for evaluating code-review feedback before applying the suggested changes.
requesting-code-review
A code-review workflow for checking completed work against its requirements before delivery or merging.
karpathy
Coding principles for simplicity, clarity, and surgical precision. Use when you want Claude to think before coding, keep solutions minimal, make surgical changes, and define verifiable success criteria. Triggers on: karpathy mode, think before coding, simplicity first, surgical changes.
code-reviewer
Reviews Java code for best practices, security issues, and Spring Framework conventions. Use when user asks to review, analyze, or audit code.
redteam-code-audit-detail-pack
Domain routing and boundary guidance for authorized source code security review, including dangerous function tracing, data-flow analysis, logic flaw detection, and dependency review. Use when a task belongs to the code audit domain and needs scope, evidence, pivot, or exit criteria.