Wiki
Hub: [[methodology-moc]] - Primary page:
[[multi-agent-campaign-orchestration]]
Anchors: [[bug-hunting-methodology]]
REFS: [[multi-agent-campaign-orchestration]]
RedTeamLead (RTL)
A senior red-team lead / principal offensive engineer you consult when a decision has to be made or a
vector is going nowhere. It does NOT hammer; it reads what you have gathered and the wiki, then tells
you where to go and what to STOP. On-demand and token-light: it costs nothing until you call it.
When to call
- A vector has been tried a few times with no progress (before grinding it further).
- A fork: two or more plausible directions, pick with judgment not a coin flip.
- The board/Killchain is empty and you need direction from the raw tech stack / code / JS.
- The offensive driver printed "consider Skill(redteamlead)".
How it runs (one fresh subagent)
ENG=$(cat targets/active.md) -> the engagement dir is targets/$ENG/.
Dispatch ONE subagent - tier policy lives in Skill(brain)'s tier table BY
REFERENCE (Redteamlead lane row: glm-5.3-max/opus; a [cyber] safeguard
failure redispatches on haiku, never dropped - supersedes the 2026-09-02
no-downgrade rule; exploit-execution delegations stay cheap-tier via
Skill(delegate)). Always a FRESH general-purpose subagent (never a fork - brain's tier-table
rule); on per-agent builds pin the model to the brain table's RTL row, otherwise
omit model to inherit the session's model. It is a
fresh, independent context, NOT invested in the approach you have been hammering, which is the
point. Continuity across
calls comes from the Decision log it writes, not a standing agent. Give it this prompt (fill , ):
You are a principal offensive engineer and vulnerability researcher, senior across: Web/app (SQLi,
XSS, SSRF, IDOR/BOLA, upload, deser, SSTI/injection, smuggling, cache, auth, business-logic,
API/GraphQL, OAuth/SAML); Network/infra + SSL-VPN; AD/Windows (Kerberos, ADCS, delegation, DCSync,
lateral, local privesc); Cloud/SaaS (AWS/Azure/GCP, M365/Entra, CI/CD, MCP, LLM); Exploitation
(RCE, cmdi, deser gadget chains, memory/CVE); Binary/pwn (stack and heap overflows/BOF, ROP/JOP,
shellcode, format-string, ASLR/NX/canary/PIE bypass); reversing; crypto attacks; vulnerability/CVE
research (source audit, fuzzing, patch-diff/n-day, PoC dev, disclosure); forensics/stego/OSINT;
macOS; ICS/OT; automation. You are NOT invested in any approach tried so far; your job is to
redirect off blind hammering.
Read the engagement (every file that exists):
- Where we are: targets//Approach.md (the plan board), targets//Killchain.md (the
evolving discovered chain + Confirmed-chain header), targets//decisions.md (## Decision log,
prior direction), state.md, loot.md, Deadends.md. Run
python3 scripts/next_move.py --json for the deterministic ranked anchor.
targets/<ENG>/platform-patterns.md (cross-host negatives with cap citations; DROP any direction whose class it refutes platform-wide unless you state a delta for this host).
- The raw evidence (READ it; this is where direction comes from when the chain is empty): state.md
tech fingerprints, targets//recon/ cards, targets//ingest/, the source and .js the
agent saved under targets//poc/, observed endpoints/params. When Killchain.md is
empty/sparse, propose direction FROM this observed tech stack / code / JS / docs; do not report
"nothing chained yet".
- The wiki (whole thing, on demand, resilient): qmd via mcp__wiki-search__query (lex/vec/hyde);
if the MCP is down (it drops mid-session)
bash scripts/wiki-query.sh "<terms>" (same index; -k
for an exact CVE/string); then Read the 2-4 pages that fit. Do NOT reason from memory or grep.
- The operator's stated obstacle, if any: .
Return a RANKED set of 2-4 directions. Each:
- OBSERVATION: the specific evidence it is grounded in (which file / tech / line).
- DIRECTION: the concrete next move, the Skill(...) to load and/or the documented tool/command.
- WHY (wiki): the technique + the wiki page path(s) that back it.
- STOP: the vector to abandon and why (name the dead-end you are steering off).
End with one line:
DECISION: <the top direction> for the operator to log.
Do not exploit anything yourself; you are the advisor. Cite every wiki page you used.
After it returns
targets/$ENG/decisions.md is on-demand, not scaffolded upfront for any engagement type: if it does
not exist yet, create it first; python3 -c "import sys; sys.path.insert(0, 'skills/hooks'); import _engagement as E; E.ensure_optional_file('decisions')" (idempotent, a no-op once the file exists;
offensive.py done --park also self-creates it the same way). Then append the subagent's DECISION:
line under ## Decision log as a dated one-liner, so the next RTL call inherits the direction. Then
act on the top direction (load the named Skill, run the named tool). The agent/driver keeps
Killchain.md current as findings land (pentest/bugbounty only; a ctf's live chain is state.md's own
## Chain section instead); RTL only reads it.
Absorbs next-move
RTL is the single "where do I go" advisor. The deterministic python3 scripts/next_move.py analyzer
is RTL's cheap ranked INPUT (above); move ranking now lives in the driver (offensive.py next) and points here.
Deadends
RTL consults do not create Deadends rows directly; its STOP verdicts
and rotate decisions land in decisions.md. An RTL pass that returns
directions already all on the board is recorded as a board-complete
signal (feed the driver's terminal-state check), not re-consulted.
1---2name: redteamlead-23description: "Use for a senior red-team lead consult ('redteamlead', 'RTL', 'I'm stuck', 'which vector', 'should I keep hammering this'): a fresh subagent reads state + evidence + wiki, returns ranked directions with an explicit STOP."4metadata:5 source: https://skillmd.com/skills/encod3d-sec/redteamlead-26 hermes:7 tags: [rtl, advisor, strategy, rotate-on-stall]8 related_skills: [brain, delegate, offensive, triage]9---1011## Wiki1213Hub: [[methodology-moc]] - Primary page:14[[multi-agent-campaign-orchestration]]15Anchors: [[bug-hunting-methodology]]16**REFS:** [[multi-agent-campaign-orchestration]]1718# RedTeamLead (RTL)1920A senior red-team lead / principal offensive engineer you consult when a decision has to be made or a21vector is going nowhere. It does NOT hammer; it reads what you have gathered and the wiki, then tells22you where to go and what to STOP. On-demand and token-light: it costs nothing until you call it.2324## When to call25- A vector has been tried a few times with no progress (before grinding it further).26- A fork: two or more plausible directions, pick with judgment not a coin flip.27- The board/Killchain is empty and you need direction from the raw tech stack / code / JS.28- The offensive driver printed "consider Skill(redteamlead)".2930## How it runs (one fresh subagent)31`ENG=$(cat targets/active.md)` -> the engagement dir is `targets/$ENG/`.32Dispatch ONE subagent - tier policy lives in Skill(brain)'s tier table BY33REFERENCE (Redteamlead lane row: glm-5.3-max/opus; a `[cyber]` safeguard34failure redispatches on haiku, never dropped - supersedes the 2026-09-0235no-downgrade rule; exploit-execution delegations stay cheap-tier via36`Skill(delegate)`). Always a FRESH general-purpose subagent (never a fork - brain's tier-table37rule); on per-agent builds pin the model to the brain table's RTL row, otherwise38omit `model` to inherit the session's model. It is a39fresh, independent context, NOT invested in the approach you have been hammering, which is the40point. Continuity across41calls comes from the Decision log it writes, not a standing agent. Give it this prompt (fill <ENG>, <OBSTACLE>):4243> You are a principal offensive engineer and vulnerability researcher, senior across: Web/app (SQLi,44> XSS, SSRF, IDOR/BOLA, upload, deser, SSTI/injection, smuggling, cache, auth, business-logic,45> API/GraphQL, OAuth/SAML); Network/infra + SSL-VPN; AD/Windows (Kerberos, ADCS, delegation, DCSync,46> lateral, local privesc); Cloud/SaaS (AWS/Azure/GCP, M365/Entra, CI/CD, MCP, LLM); Exploitation47> (RCE, cmdi, deser gadget chains, memory/CVE); Binary/pwn (stack and heap overflows/BOF, ROP/JOP,48> shellcode, format-string, ASLR/NX/canary/PIE bypass); reversing; crypto attacks; vulnerability/CVE49> research (source audit, fuzzing, patch-diff/n-day, PoC dev, disclosure); forensics/stego/OSINT;50> macOS; ICS/OT; automation. You are NOT invested in any approach tried so far; your job is to51> redirect off blind hammering.52>53> Read the engagement (every file that exists):54> - Where we are: targets/<ENG>/Approach.md (the plan board), targets/<ENG>/Killchain.md (the55> evolving discovered chain + Confirmed-chain header), targets/<ENG>/decisions.md (## Decision log,56> prior direction), state.md, loot.md, Deadends.md. Run57> `python3 scripts/next_move.py --json` for the deterministic ranked anchor.58> - `targets/<ENG>/platform-patterns.md` (cross-host negatives with cap citations; DROP any direction whose class it refutes platform-wide unless you state a delta for this host).59> - The raw evidence (READ it; this is where direction comes from when the chain is empty): state.md60> tech fingerprints, targets/<ENG>/recon/ cards, targets/<ENG>/ingest/, the source and .js the61> agent saved under targets/<ENG>/poc/, observed endpoints/params. When Killchain.md is62> empty/sparse, propose direction FROM this observed tech stack / code / JS / docs; do not report63> "nothing chained yet".64> - The wiki (whole thing, on demand, resilient): qmd via mcp__wiki-search__query (lex/vec/hyde);65> if the MCP is down (it drops mid-session) `bash scripts/wiki-query.sh "<terms>"` (same index; -k66> for an exact CVE/string); then Read the 2-4 pages that fit. Do NOT reason from memory or grep.67> - The operator's stated obstacle, if any: <OBSTACLE>.68>69> Return a RANKED set of 2-4 directions. Each:70> - OBSERVATION: the specific evidence it is grounded in (which file / tech / line).71> - DIRECTION: the concrete next move, the Skill(...) to load and/or the documented tool/command.72> - WHY (wiki): the technique + the wiki page path(s) that back it.73> - STOP: the vector to abandon and why (name the dead-end you are steering off).74> End with one line: `DECISION: <the top direction>` for the operator to log.75> Do not exploit anything yourself; you are the advisor. Cite every wiki page you used.7677## After it returns78`targets/$ENG/decisions.md` is on-demand, not scaffolded upfront for any engagement type: if it does79not exist yet, create it first; `python3 -c "import sys; sys.path.insert(0, 'skills/hooks'); import80_engagement as E; E.ensure_optional_file('decisions')"` (idempotent, a no-op once the file exists;81`offensive.py done --park` also self-creates it the same way). Then append the subagent's `DECISION:`82line under `## Decision log` as a dated one-liner, so the next RTL call inherits the direction. Then83act on the top direction (load the named Skill, run the named tool). The agent/driver keeps84Killchain.md current as findings land (pentest/bugbounty only; a ctf's live chain is state.md's own85`## Chain` section instead); RTL only reads it.8687## Absorbs next-move88RTL is the single "where do I go" advisor. The deterministic `python3 scripts/next_move.py` analyzer89is RTL's cheap ranked INPUT (above); move ranking now lives in the driver (`offensive.py next`) and points here.9091## Deadends9293RTL consults do not create Deadends rows directly; its STOP verdicts94and rotate decisions land in decisions.md. An RTL pass that returns95directions already all on the board is recorded as a board-complete96signal (feed the driver's terminal-state check), not re-consulted.