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 excessive-agencygit 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/excessive-agency)<a href="https://agentmods.dev/skills/purpleailab/decepticon/excessive-agency"><img src="https://agentmods.dev/badge/skills/purpleailab/decepticon/excessive-agency/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/excessive-agency"><img src="https://agentmods.dev/badge/skills/purpleailab/decepticon/excessive-agency.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector warn
SkillSpector: 2 findings, 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 Excessive Agency · line 27 Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.Fix: Add human-in-the-loop confirmation for destructive, irreversible, or high-impact operations. Never auto-execute commands that modify files, send data, or alter system state.
- medium Excessive Agency · line 42 Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.Fix: Add human-in-the-loop confirmation for destructive, irreversible, or high-impact operations. Never auto-execute commands that modify files, send data, or alter system state.
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 | $0.00067 | $0.01449 |
| Opus 5 | $0.00034 | $0.00724 |
| Sonnet 5 | $0.00013 | $0.00290 |
| Haiku 4.5 | $0.00007 | $0.00145 |
Grade A, and why
excessive-agency 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 — 144 lines — stays where its author put it; the contents beside it link to each section on GitHub.
LLM Excessive Agency (LLM06:2025)
A model that can send_email can mass-mail customers; a model that
can execute_sql can drop tables; a model that can both read_inbox
and send_email is a data-exfiltration primitive in 20 lines. The
vuln is not the individual tool — it's the combination, the
permission scope, and the lack of approval gates. Excessive
agency frequently weaponises an LLM01 prompt injection into business-
material impact.
1. Recognition signals
- "AI assistant" auto-acts on user data (calendar, email, files, repos).
- Tool list contains anything that writes or calls outbound:
send_*,delete_*,execute_*,payment_*,deploy_*. - Tool wrappers do not require human confirmation for destructive ops.
- Tools authorised at install time with broad scopes (
read+write+admin). - One service account with all permissions, used by every tool.
- Auto-approve flag (e.g.
--yes/approve_all) wired by default. - Long-running agent loops with no per-step budget.
2. Attack vectors
Excessive functionality
Tools exist that the use case doesn't need — a customer-support bot
with execute_terraform, a sales co-pilot with read_payroll.
Even unused tools become attack surface (LLM01 picks one).
Excessive permissions per tool
send_email accepts arbitrary recipients including external. The
underlying SMTP creds let it relay anywhere.
Excessive autonomy
No human-in-the-loop. payment_* runs without confirmation.
delete_repository reachable from chat. deploy_to_prod callable
on a single tool call.
Shared identity
Every tool uses the same service account with workspace-admin rights. A compromise of one capability gives all of them.
Persistent memory + agentic loop
The model decides multi-step plans and executes without re-asking. Once injected, the loop completes the attacker's plan without ever re-prompting the user.
3. Audit workflow
# Enumerate the tool inventory
grep -rE '@tool|tools\s*=' /workspace/src -A 3 | head -200
# Find destructive verbs in tool definitions
grep -rEi '(send|delete|drop|execute|deploy|pay|wire|transfer|grant|revoke)_' /workspace/src
# Find approval / confirm logic adjacent to tool calls (or its absence)
grep -rEi 'require.*confirm|human_in_loop|approve|interrupt_before' /workspace/src
# Find service-account creds tied to tools
grep -rE 'service_account|admin_token|workspace_admin|SUPER_USER|elevated' /workspace/src
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 · 144 lines · 67 tokens per session scan A d1298d77ee17
excessive-agency is a skill published in the GitHub repository PurpleAILAB/Decepticon (5,491 stars, last pushed 13d ago), licensed Apache-2.0. It adds 67 tokens to every session and 1,449 once invoked, about $0.0003 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 2026-09-03.
Other skills, from other repositories
interactive-dashboard
Interactive web dashboards: stock trackers, sector heatmaps, portfolio monitors — served via preview URL.
onboarding
First-time user onboarding to set up investment profile, watchlists, portfolio, and preferences.
idea-generation
Stock screening and idea generation: quantitative screens, thematic analysis, shortlist.
secretary
Workspace and research management — dispatch analyses, monitor running agents, manage workspaces and threads.
python-lib-analyzer
Analyze any Python library structure, explore modules, classes, and functions with signatures and documentation.
analyzing-windows-prefetch-with-python
Use when parse Windows Prefetch files using the windowsprefetch Python library to reconstruct application execution history, detect renamed or masquerading binaries, and identify suspicious program execution patterns. Use when working with analyzing windows prefetch with python.