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 JesusConwellpy/secagent-skills --skill secagent-sandboxgit clone --depth 1 https://github.com/JesusConwellpy/secagent-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/jesusconwellpy/secagent-skills/secagent-sandbox)<a href="https://agentmods.dev/skills/jesusconwellpy/secagent-skills/secagent-sandbox"><img src="https://agentmods.dev/badge/skills/jesusconwellpy/secagent-skills/secagent-sandbox/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/jesusconwellpy/secagent-skills/secagent-sandbox"><img src="https://agentmods.dev/badge/skills/jesusconwellpy/secagent-skills/secagent-sandbox.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 | $0.00774 |
| Opus 5 | $0.00011 | $0.00387 |
| Sonnet 5 | $0.00004 | $0.00155 |
| Haiku 4.5 | $0.00002 | $0.00077 |
Grade B, and why
secagent-sandbox scanned grade B 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 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.
Downloads and executes remote codemediumSupply chain
curl | sh runs whatever the server returns today, which is not necessarily what it returned when this was reviewed.
curl ... | bash (pipe to shell) Downgraded: this mod is about security review, or the phrase is quoted, so it is likely naming the pattern rather than instructing it.
Makes network callslowCapability
Not a fault in itself. Listed so you know the mod talks to something, and to what.
curl GET (read network) → LOW (auto-approved, authorized targets only) How it starts
The opening of the file, as written. The whole thing — 96 lines — stays where its author put it; the contents beside it link to each section on GitHub.
SecAgent Sandbox
Loading this SKILL enforces a security protocol before any tool execution.
Authorization Check (First step — mandatory)
1. Does the target match *.ctfhub.com, *.hackthebox.com, *.tryhackme.com?
→ YES: CTF-exempt. Proceed.
2. Is the target localhost/127.0.0.1?
→ YES: Proceed.
3. Check ~/.secagent/authorization.toml:
- CIDR match? → Proceed.
- Domain wildcard match? → Proceed.
- File doesn't exist? → Create it (empty file = deny-all except localhost + CTF).
4. None of the above?
→ REJECT. Output: "Target {target} not authorized."
authorization.toml Bootstrap
If ~/.secagent/authorization.toml doesn't exist:
mkdir -p ~/.secagent
touch ~/.secagent/authorization.toml
Default content:
# SecAgent Authorization
# Add authorized targets below.
# CIDR example: 192.168.0.0/16
# Domain example: *.example.com
# CTF platforms (*.ctfhub.com, *.hackthebox.com, *.tryhackme.com) are auto-authorized.
Default behavior: empty file → allow only localhost + CTF platforms.
Operation Risk Levels
Read file (cat, head, read) → LOW (auto-approved)
curl GET (read network) → LOW (auto-approved, authorized targets only)
Write file (write, edit) → MEDIUM (auto-approved inside workspace)
nmap (network scan) → MEDIUM (authorized targets only)
Execute PoC (exploit) → HIGH (allowed, logged)
Dangerous command → CRITICAL (rejected or reported)
Dangerous patterns (immediate rejection):
bash -c "..." (inline command execution)
curl ... | bash (pipe to shell)
$(...) or `` (command substitution)
/dev/tcp (reverse shell)
nc -e / ncat -e (netcat reverse shell)
>/dev/null 2>&1 with dangerous commands (output suppression)
Audit Log
Log one line per operation:
[EXEC] {ISO8601_UTC} | {tool} | exit={code} | {duration_ms}ms | {risk}
ISO8601_UTC: 2026-05-25T10:30:00Z format, always UTC
exit_code: 0-255, or BLOCKED, or TIMEOUT
duration_ms: total time from assessment to completion; assessment time if blocked
risk: LOW / MEDIUM / HIGH / CRITICAL
Examples:
[EXEC] 2026-05-25T10:30:01Z | curl | exit=0 | 230ms | LOW
[EXEC] 2026-05-25T10:31:00Z | curl | exit=BLOCKED | 5ms | HIGH
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 · 96 lines · 22 tokens per session scan B aea22c31bd0f
secagent-sandbox is a skill published in the GitHub repository JesusConwellpy/secagent-skills (2 stars, last pushed 3mo ago), licensed MIT. It adds 22 tokens to every session and 774 once invoked, about $0.0001 per session on Opus 5. A static security scan graded it B with 2 findings (downloads and executes remote code, makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
osint-recon
A knowledge guide for open-source intelligence, or OSINT: collecting information from publicly available sources. It covers servers, websites, domains, and, when appropriate, people.
secknowledge-skill
A knowledge base for testing the security of websites, software, and AI systems, including agents and language models.
performing-ai-assisted-vulnerability-discovery
Using LLMs to accelerate vulnerability research and pentest workflows — generating syntax-valid fuzzing seeds and evolving grammars, fine-tuned mutation dictionaries, parallel agent-based proof-of-vulnerability generation, and evidence-driven passive analysis of real HTTP traffic via the Burp MCP server. Covers…
client-reverse
A workflow for analysing and replaying requests made by mobile apps, browser JavaScript, or desktop clients during authorised security testing.
redteam-crypto-detail-pack
Domain routing and boundary guidance for authorized cryptography weakness testing, including weak algorithms, padding oracles, key management errors, insecure randomness, and hash collision risks. Use when a task belongs to the cryptography testing domain and needs scope, evidence, pivot, or exit criteria.
redteam-logic-detail-pack
Domain routing and boundary guidance for authorized business logic vulnerability testing, including race conditions, flow bypass, price tampering, permission logic errors, and bulk operation abuse. Use when a task belongs to the logic testing domain and needs scope, evidence, pivot, or exit criteria.