wiki-arsenal
The fast, wiki-first lookup engine for "what do I use against this surface". Runs the four
knowledge areas in parallel so a deep lookup is one wall-clock, not four serial reads. The
hunt-* skills each carry their own wiki-first mcp__wiki-search__query (MCP-independent) and
can hand off here for a fast parallel lookup. Never hand-roll from memory when the wiki has the answer.
Input: a surface, service, or vuln-class (e.g. Jenkins on 8080, SSRF, Kerberoasting).
Wiki
Hub: [[methodology-moc]] - Primary page: [[bug-hunting-methodology]]
Anchors: [[vuln-assessment]]
REFS: [[bug-hunting-methodology]]
0. Cache check first (0 tokens on a repeat)
Slug the surface (lowercase, non-alnum -> -). If targets/<active-eng>/arsenal/<slug>.md
exists, read and return it. Do not re-spend. (<active-eng> = the dir named in targets/active.md.)
Mode: quick (DEFAULT - fire it constantly)
One mcp__wiki-search__query over the whole index (add a mcp__wiki-search__query (lex) keyword pass when the
surface is an exact product/CVE string). Group the hits under the four areas and return each as
path -> one-line snippet:
- Techniques (
wiki/techniques/)
- Payloads (
wiki/payloads/)
- Tools (
wiki/tools/)
- Cheatsheets (
wiki/cheatsheets/)
Cost ~1-2k tokens, no subagents. Fire it on
every new surface to raise wiki coverage cheaply. Stop here unless the surface is worth deep prep.
Volume: quick-arsenal arms on a genuinely new surface, not per action - the step 0 cache covers
repeats, and per-action lookups ride the hunt skill's own reference map (hunt-core's tiered wiki lookup).
Mode: deep (opt-in - "deep"/"full arsenal", or a whole service/target worth prepping)
Dispatch FOUR parallel subagents in a SINGLE message (Agent tool), one per area, with
on the cheap tier (glm-5.3-flash; pin model only if your seat build exposes per-agent selection) - each only
reads its area and distils a card, which a lightweight model does well at a fraction of the cost (a
full-model fan-out measured ~170k tokens; the cheap tier cuts that hard). Each
is told to search only its area, read the top 2-3 matching pages, and return a compact ready-to-use
card for its area ONLY (nothing else), citing the page paths it used:
| Agent |
Searches |
Returns |
| tools |
wiki/tools/ |
the automated tool(s) to run + the exact command line |
| payloads |
wiki/payloads/ |
ready-to-send payloads for the vuln-class |
| techniques |
wiki/techniques/ |
the attack steps / chain |
| cheatsheets |
wiki/cheatsheets/ |
quick copy-paste commands |
Each agent scopes its search to its area: pass a path filter to qmd, or query the whole index
and keep only wiki/<area>/ hits, then read those pages. The pages an agent reads stay in that
agent's context and are discarded; you ingest only its card.
Merge the four returned cards into one arsenal card, four labelled sections plus a final
Sources: line listing every page used.
Persist the deep card
Write the merged card to targets/<active-eng>/arsenal/<slug>.md (create the dir on demand) with
frontmatter surface: and generated:. That is the cache step 0 reads next time.
Guardrails (token control)
- Default is quick (~1-2k tokens). Only go deep on request or for a real service/target.
- Deep is bounded: exactly 4 agents on the cheap tier, each capped to the top 2-3 pages. The cost
is isolated to the subagents; your main context only gains the merged card.
- The cache prevents re-spend on the same surface.
Hand off
The full class-specific methodology lives in the matching hunt-* skill. After the arsenal card,
hand off to it (e.g. Skill(hunt-ssrf)) for the actual exploitation loop.
At-a-glance index (folded in from the retired arsenal skill)
Sanity-check the lookup against these tables; they are not a substitute for the qmd pass. Order:
automated tool -> technique/payload -> capture.
Tool first (wiki/tools/ - don't improvise)
Fingerprint the surface -> reach for the tool we already document. ls wiki/tools/ for all of them.
| Surface / service |
Automated tools (wiki/tools/<name>.md) |
| Web HTTP(S) |
httpx, whatweb, nikto -> ffuf/feroxbuster/gobuster (content) -> katana/gau (crawl) -> arjun (params) -> nuclei (CVE/misconfig) -> dalfox (XSS) -> wpscan (WP); gowitness, burp-suite/burp-mcp |
| Ports / host |
nmap, rustscan, naabu |
| DNS / subdomains |
subfinder, amass, dnsx, gau |
| SMB / Windows / AD |
netexec, responder, impacket, bloodhound/powerview, kerbrute, certipy/rubeus, evil-winrm |
| Login / creds |
hydra, medusa (brute) -> hashcat, john (crack) -> jwt_tool (JWT) -> swaks (SMTP) |
| Post-shell privesc |
pspy + linpeas/peass (ALWAYS, first) |
| Pivot / tunnel |
chisel, ligolo-ng |
| Cloud |
pacu, scoutsuite, roadtools, trivy |
| Binary / RE / pwn |
ghidra, radare2, gdb-gef, pwntools, angr, binwalk, jadx, apktool, frida |
| Secrets / SAST / forensics |
trufflehog, trivy, semgrep, codeql, volatility, tshark |
Read the tool page for exact flags before running; no page -> mcp__wiki-search__query/Skill(wiki).
Then technique / payload (wiki/payloads/ + wiki/cheatsheets/)
- Vuln class ->
wiki/payloads/<class>.md: sqli, xss, ssrf, ssti, xxe, idor, nosql, jwt,
deserialization, command-injection, lfi-path-traversal, csrf, cors, crlf, graphql, api,
auth-bypass, session, race-conditions, prototype-pollution, open-redirect, host-header,
smuggling, web-cache, oauth-saml, mfa-bypass, crypto, ldap, xpath, webauthn-passkey,
file-upload, imds-cloud-metadata, llm-prompt-injection, modbus, cicd.
- Exploitation / privesc / chains ->
wiki/cheatsheets/*: privesc-exploit-arsenal,
cve-arsenal, attack-chains, linux-privesc / windows-privesc, default-credentials,
nuclei-arsenal, sqlmap, password-attacks.
Read the matching page(s) BEFORE hand-rolling a payload/exploit. The full class-specific
methodology lives in the matching hunt-* skill - hand off to it.
1---2name: wiki-arsenal-23description: Parallel wiki lookup for a surface/vuln-class across techniques/payloads/tools/cheatsheets; quick (one qmd search) or deep (4 parallel subagents). Use for "what do I use against X", "arsenal for X", "deep/full arsenal".4metadata:5 source: https://skillmd.com/skills/encod3d-sec/wiki-arsenal-26 hermes:7 tags: [wiki, arsenal, tooling-lookup]8 related_skills: [wiki, hunt-core]9---1011# wiki-arsenal1213The fast, wiki-first lookup engine for "what do I use against this surface". Runs the four14knowledge areas in parallel so a deep lookup is one wall-clock, not four serial reads. The15`hunt-*` skills each carry their own wiki-first `mcp__wiki-search__query` (MCP-independent) and16can hand off here for a fast parallel lookup. Never hand-roll from memory when the wiki has the answer.1718Input: a surface, service, or vuln-class (e.g. `Jenkins on 8080`, `SSRF`, `Kerberoasting`).1920## Wiki2122Hub: [[methodology-moc]] - Primary page: [[bug-hunting-methodology]]23Anchors: [[vuln-assessment]]24**REFS:** [[bug-hunting-methodology]]2526## 0. Cache check first (0 tokens on a repeat)2728Slug the surface (lowercase, non-alnum -> `-`). If `targets/<active-eng>/arsenal/<slug>.md`29exists, read and return it. Do not re-spend. (`<active-eng>` = the dir named in `targets/active.md`.)3031## Mode: quick (DEFAULT - fire it constantly)3233One `mcp__wiki-search__query` over the whole index (add a `mcp__wiki-search__query (lex)` keyword pass when the34surface is an exact product/CVE string). Group the hits under the four areas and return each as35`path -> one-line snippet`:3637- **Techniques** (`wiki/techniques/`)38- **Payloads** (`wiki/payloads/`)39- **Tools** (`wiki/tools/`)40- **Cheatsheets** (`wiki/cheatsheets/`)4142Cost ~1-2k tokens, no subagents. Fire it on43every new surface to raise wiki coverage cheaply. Stop here unless the surface is worth deep prep.44Volume: quick-arsenal arms on a genuinely new surface, not per action - the step 0 cache covers45repeats, and per-action lookups ride the hunt skill's own reference map (hunt-core's tiered wiki lookup).4647## Mode: deep (opt-in - "deep"/"full arsenal", or a whole service/target worth prepping)4849Dispatch FOUR parallel subagents in a SINGLE message (Agent tool), one per area, with50on the cheap tier (glm-5.3-flash; pin `model` only if your seat build exposes per-agent selection) - each only51reads its area and distils a card, which a lightweight model does well at a fraction of the cost (a52full-model fan-out measured ~170k tokens; the cheap tier cuts that hard). Each53is told to search only its area, read the top 2-3 matching pages, and return a compact ready-to-use54card for its area ONLY (nothing else), citing the page paths it used:5556| Agent | Searches | Returns |57|---|---|---|58| tools | `wiki/tools/` | the automated tool(s) to run + the exact command line |59| payloads | `wiki/payloads/` | ready-to-send payloads for the vuln-class |60| techniques | `wiki/techniques/` | the attack steps / chain |61| cheatsheets | `wiki/cheatsheets/` | quick copy-paste commands |6263Each agent scopes its search to its area: pass a path filter to `qmd`, or query the whole index64and keep only `wiki/<area>/` hits, then read those pages. The pages an agent reads stay in that65agent's context and are discarded; you ingest only its card.6667Merge the four returned cards into one arsenal card, four labelled sections plus a final68`Sources:` line listing every page used.6970### Persist the deep card7172Write the merged card to `targets/<active-eng>/arsenal/<slug>.md` (create the dir on demand) with73frontmatter `surface:` and `generated:`. That is the cache step 0 reads next time.7475## Guardrails (token control)7677- Default is quick (~1-2k tokens). Only go deep on request or for a real service/target.78- Deep is bounded: exactly 4 agents on the cheap tier, each capped to the top 2-3 pages. The cost79 is isolated to the subagents; your main context only gains the merged card.80- The cache prevents re-spend on the same surface.8182## Hand off8384The full class-specific methodology lives in the matching `hunt-*` skill. After the arsenal card,85hand off to it (e.g. `Skill(hunt-ssrf)`) for the actual exploitation loop.8687## At-a-glance index (folded in from the retired arsenal skill)8889Sanity-check the lookup against these tables; they are not a substitute for the qmd pass. Order:90automated tool -> technique/payload -> capture.9192### Tool first (`wiki/tools/` - don't improvise)9394Fingerprint the surface -> reach for the tool we already document. `ls wiki/tools/` for all of them.9596| Surface / service | Automated tools (`wiki/tools/<name>.md`) |97|---|---|98| Web HTTP(S) | httpx, whatweb, nikto -> ffuf/feroxbuster/gobuster (content) -> katana/gau (crawl) -> arjun (params) -> nuclei (CVE/misconfig) -> dalfox (XSS) -> wpscan (WP); gowitness, burp-suite/burp-mcp |99| Ports / host | nmap, rustscan, naabu |100| DNS / subdomains | subfinder, amass, dnsx, gau |101| SMB / Windows / AD | netexec, responder, impacket, bloodhound/powerview, kerbrute, certipy/rubeus, evil-winrm |102| Login / creds | hydra, medusa (brute) -> hashcat, john (crack) -> jwt_tool (JWT) -> swaks (SMTP) |103| Post-shell privesc | pspy + linpeas/peass (ALWAYS, first) |104| Pivot / tunnel | chisel, ligolo-ng |105| Cloud | pacu, scoutsuite, roadtools, trivy |106| Binary / RE / pwn | ghidra, radare2, gdb-gef, pwntools, angr, binwalk, jadx, apktool, frida |107| Secrets / SAST / forensics | trufflehog, trivy, semgrep, codeql, volatility, tshark |108109Read the tool page for exact flags before running; no page -> `mcp__wiki-search__query`/`Skill(wiki)`.110111### Then technique / payload (`wiki/payloads/` + `wiki/cheatsheets/`)112113- **Vuln class -> `wiki/payloads/<class>.md`**: sqli, xss, ssrf, ssti, xxe, idor, nosql, jwt,114 deserialization, command-injection, lfi-path-traversal, csrf, cors, crlf, graphql, api,115 auth-bypass, session, race-conditions, prototype-pollution, open-redirect, host-header,116 smuggling, web-cache, oauth-saml, mfa-bypass, crypto, ldap, xpath, webauthn-passkey,117 file-upload, imds-cloud-metadata, llm-prompt-injection, modbus, cicd.118- **Exploitation / privesc / chains -> `wiki/cheatsheets/*`**: privesc-exploit-arsenal,119 cve-arsenal, attack-chains, linux-privesc / windows-privesc, default-credentials,120 nuclei-arsenal, sqlmap, password-attacks.121122Read the matching page(s) BEFORE hand-rolling a payload/exploit. The full class-specific123methodology lives in the matching `hunt-*` skill - hand off to it.